Jrockit Mission Control: Oracle Jrockit Runtime Analyzer

BEA JRockit ®
Mission
Control TM
Oracle JRockit Runtime

Analyzer
JRockit Mission Control version 3.0.2 ®

Document Revised: June, 2008
Contents
Introduction to the Oracle JRockit Runtime Analyzer

How Does the JRockit Runtime Analyzer Work? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-1
What is a JRA Recording? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-2
What is the JRA Tool? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-2
What’s New in the JRA System? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-3
Workflow Description for Creating and Analyzing a JRA

Recording
1. Start the JRockit Mission Control Client . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-2
2. Start Your Java Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-2
3. Create JRA Recording. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-3
4. View Your JRA Recording in JRA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-8
5. Perform Changes in Application or Use Other Command-line Options for the JRockit
JVM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-9
6. Create a New JRA Recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-10
7. Compare and Contrast Two Recordings in JRA . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-10
Alternative Ways to Start a JRA Recording

Starting a Recording with jrcmd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-1
Starting a Recording From the Command Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-2
Oracle JRockit Runtime Analyzer iii

Using the JRockit Runtime Analyzer
Getting Started with the JRockit Runtime Analyzer
Starting JRA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-1
JRA Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-1
JRA Tabs at a Glance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-3
Customizing Your JRA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-3
Turning on/off Tabs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-4
Changing Table Settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-5
Filtering Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-6
Collapsing and Expanding an Information Panel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-7
Changing Layout of a Tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-7
General Information in a JRA Recording

Getting Familiar with the General Tab. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-2
Viewing General Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-3
Viewing Memory Usage Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-4
Viewing Miscellaneous Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-4
Viewing Memory Allocation Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-6
Viewing Threads Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-7
Viewing Exceptions Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-8
Methods and Call Trace Information

Getting Familiar with the Methods Tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-1
Viewing Hot Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-3
Viewing Predecessors and Successors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-3
Jumping to Application Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-4
iv Oracle JRockit Runtime Analyzer

General Garbage Collector Information
Getting Familiar with the GC General Tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-1
Viewing General Garbage Collection Information. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-2
Viewing Garbage Collection Call Tree Information . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-3
Viewing Garbage Collection Strategy Changes Information . . . . . . . . . . . . . . . . . . . . . 8-3
Jumping to Application Source. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-4
Garbage Collection Events Information

Getting Familiar with the GCs Tab. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-2
Changing Focus on GC Chart. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-3
Viewing Specifics about Garbage Collections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-4
Viewing the Detailed Information About the Garbage Collection . . . . . . . . . . . . . . . . . 9-6
Viewing Information on the Common Garbage Collection Tab . . . . . . . . . . . . . . . . . . 9-7
Viewing Information on the GC Method Call Tree Tab . . . . . . . . . . . . . . . . . . . . . . . . 9-9
Viewing Information on the Old/Young Collection Tab . . . . . . . . . . . . . . . . . . . . . . . . 9-9
Viewing Information on the Cache Lists Tab. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-10
The Pause Time Tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-11
Jumping to Application Source. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-11
Java Heap Content Information

Getting Familiar with the Heap Tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-1
Viewing the Heap Snapshot at the End of the Recording Information . . . . . . . . . . . . 10-2
Viewing the Heap Contents Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-3
Viewing the Free Memory Contribution Information . . . . . . . . . . . . . . . . . . . . . . . . . 10-3
Objects Information
Getting Familiar with the Objects Tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-1
Viewing Start of Recording Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-2
Viewing End of Recording Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-3
Oracle JRockit Runtime Analyzer v

Jumping to Application Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-3
Code Optimization Information

Getting Familiar with the Optimizations Tab. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-1
Viewing Optimization Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-2
Viewing Methods Optimized During Recording Information . . . . . . . . . . . . . . . . . . . 12-3
Lock Profiling Information

Getting Familiar with the Locks Tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-1
Java Locks Profiling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-2
Native Lock Profiling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-4
Start and End Processes Information

Turning on the Processes Tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-1
Getting Familiar with the Processes Tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-2
Snapshot of Processes at Beginning and End of Recording. . . . . . . . . . . . . . . . . . . . . 14-3
Detailed Processes Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-4
Threads Information
Turning on the Threads Tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15-1
Getting Familiar with the Threads Tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15-2
List of Times when Thread Dump is Taken. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15-3
Thread Dump Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15-4
Using the Latency Tabs

Latency Tabs at a Glance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-2
Creating a JRA Recording with Latency Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-3
Opening a JRA Recording that Contains Latency Data . . . . . . . . . . . . . . . . . . . . . . . . 16-7
vi Oracle JRockit Runtime Analyzer

Shared Functionality Amongst All Latency Views . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-8
Using the Latency Timeline Slide Bar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-9
What is an Operative Set?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-11
Working with an Operative Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-12
About the Event Types View . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-13
Using the Event Types View to Decrease Displayed Events . . . . . . . . . . . . . . . . . . . 16-14
Using the Event Types View to Work with Operative Sets . . . . . . . . . . . . . . . . . . . . 16-15
About the Properties View . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-16
Example of How to Compare two JRA Recordings where one Contains Latencies . 16-17
Example Workflow of How to Find Latencies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-19
1. Create a JRA Recording with Latency Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-20
2. Open the JRA Recording in the Latency Graph Tab . . . . . . . . . . . . . . . . . . . . . . . 16-21
3. Look on the Latency Traces Tab to Find Specific Method . . . . . . . . . . . . . . . . . . 16-21
4. Add a Suspected Method to the Operative Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-22
5. Look at Operative Set on the Latency Traces Tab . . . . . . . . . . . . . . . . . . . . . . . . . 16-23
6. Perform Changes to Your Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-24
7. Compare and Contrast Recordings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-24
Filtering Latency Event Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-25
Latency Log Information

Getting Familiar with the Latency Log Tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-2
Changing Start Time View on an Event . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-3
About Details for Events. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-4
Selecting an Event . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-4
Understanding Event Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-5
Viewing General Event Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-6
Viewing Event Property Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-7
Viewing Event Stack Traces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-7
Oracle JRockit Runtime Analyzer vii

Viewing Event Property Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-8
Latency Graph Information

Getting Familiar with the Latency Graph Tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18-1
Using the Latency Timeline Slide Bar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18-3
Understanding the Different Parts of a Thread Image . . . . . . . . . . . . . . . . . . . . . . . . . 18-3
Filtering on Thread Names. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18-4
What Does the Threads Chart Contain? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18-5
Correlating Events on Threads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18-5
Magnifying a Thread . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18-6
Showing Thread Transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18-7
Hovering Over an Event. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18-8
Showing Garbage Collection Backdrop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18-9
Latency Traces Information

Getting Familiar with the Latency Traces Tab. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19-1
Setting Trace Filter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19-3
Adding Comments and Notes to a Recording
viii Oracle JRockit Runtime Analyzer

CHAPTER 1
Introduction to the Oracle JRockit

Runtime Analyzer
The JRockit Runtime Analyzer (JRA) is a Java application and JVM profiler that is especially
designed for the Oracle JRockit JVM. JRA is a well integrated part of Oracle JRockit Mission
Control and measures performance in a non-intrusive way in both production and development
environments.
This section is divided into the following topics:
z How Does the JRockit Runtime Analyzer Work?
z What is a JRA Recording?
z What is the JRA Tool?
z What’s New in the JRA System?
How Does the JRockit Runtime Analyzer Work?

The JRA consists of two components (Figure 1-1): one component inside the JRockit JVM that
collects and saves data (the JRA recording engine); and an analysis tool that visualizes the
information (JRA). The JRockit JVM-internal component produces a recording of the system’s
runtime behavior during a user specified period of time, typically a few minutes. The recording
results in an XML file that is automatically transferred to the JRockit Mission Control Client and
opened in the JRA tool (this behavior is valid for JRockit JDK 5.0 and later; for JRockit JDK 1.4,
the file is saved locally to disk and you need to locate it before opening the file).
Oracle JRockit Runtime Analyzer 1-1

I n tr o d uct i o n t o t h e O r acl e J R o c k i t R un ti me A na l y z e r
The recording is a great way to share how the JRockit JVM has worked with your application.
You can also use several recordings to compare and contrast how different command line options
change the behavior of your application, for example, by creating before-and-after recordings.
When sending trouble reports to the JRockit support team, you are required to attach a JRA
recording to your trouble report. The recording is analyzed “offline” by the JRA.
Figure 1-1 The JRockit Runtime Analyzer
The recording engine uses several sources of information including the JRockit JVM hot spot
detector (also used by the optimization engine to decide what methods to optimize), the operating
system, the JRockit JVM memory system (most notably the garbage collector), the JRockit JVM
thread analyzer (if enabled), and the JRockit JVM lock profiler (if enabled).
What is a JRA Recording?

The JRA recording is a collection of data about the JVM and the running Java application. This
recording can be used in the JRA to analyze what happened in the JRockit JVM and the Java
application itself.
What is the JRA Tool?

The JRA tool parses a JRA recording and visualizes the data. This is a convenient way to analyze
the data offline. The size of the compressed recording is on the order of a few hundred kilobytes,
so a system administrator can easily make a recording of a deployed system and send it to the
JVM or application developer who probably is in a better position to analyze it.
JRA shows a top list of the hottest methods where you can select a method and see its call tree,
i.e. its predecessors (what other methods have called this method) and successors (what methods
the selected method will call). A percentage for each branch indicates how common a given path
is.
1-2 Oracle JRockit Runtime Analyzer

Wh at ’ s N e w i n t he J R A S y s t e m ?
As for memory management, there is a graph of the varying heap usage and pause times for the
garbage collections. Detailed information about each GC shows exactly how much memory was
released in a collection. There are also pie charts showing the distributions in size of free memory
blocks and the distribution of occupied memory in small and large object chunks.
What’s New in the JRA System?

In Oracle JRockit Mission Control 3.0, JRA has been extended to record even more information
about your Java application and about the JVM itself. The JRA engine now has the possibility
record thread related information and JRA has been extended with new tabs to visualize thread
and thread latency information.
Another nifty feature that will make your JRA work space less cluttered, is that you can turn off
tabs in JRA that are not showing any data. If you are a returning user to JRA, you will also find
that the recording dialog for JRA gives you more possibilities to take control over the recording
itself, which data to include and not include, etc.
The thread and thread latency feature can help you pinpoint, down to the method, where you
might have bottlenecks or problems in the application.

I n tr o d uct i o n t o t h e O r acl e J R o c k i t R un ti me A na l y z e r

CHAPTER 2
Workflow Description for Creating and

Analyzing a JRA Recording
This section is a workflow description of how to use the JRA system to find problems and
improvement areas with your Java application and the Oracle JRockit JVM. JRA is excellent to
use when tuning your system, for example when looking for performance bottlenecks, such as
latencies. The typical workflow when working with JRA is described in Figure 2-1.
Figure 2-1 Typical workflow for comparing different JRockit JVM and Java application settings
The first steps are to start the JRockit Mission Control Client and then start you application so
that you can start a JRA recording. The JRA recording takes a snapshot of the system’s runtime
behavior during the time period that you specify, typically a few minutes. As soon as the
recording is complete, it opens in JRA where it can be analyzed “offline”. If you want to, you can

Wo rkf l o w D e s c r i pt i o n f o r C r e at in g an d A na l y z i n g a J R A R e c o r di ng
perform changes to your application or change command-line options for the JRockit JVM and
create a new recording. This way, you have a chance to compare and contrast how different
settings affect your application.
The steps for creating and comparing and contrasting a JRA recording are detailed in the
following topics:
1. Start the JRockit Mission Control Client
2. Start Your Java Application
3. Create JRA Recording
4. View Your JRA Recording in JRA
5. Perform Changes in Application or Use Other Command-line Options for the JRockit
JVM
6. Create a New JRA Recording
7. Compare and Contrast Two Recordings in JRA
1. Start the JRockit Mission Control Client

The way you start the JRockit Mission Control Client depends on which platform you are running
it on.
Windows platforms:
Click Start > All Programs > Oracle JRockit Mission Control 3.0.3 for Java SE <JDK
version> > Oracle JRockit Mission Control 3.0.3 or invoke the launcher
(JROCKIT_HOME\bin\jrmc.exe).
Unix platforms:
JROCKIT_HOME/bin/jrmc
2. Start Your Java Application

1. Start your Java application with the JRockit JVM.
– If you are running the JRockit JVM 1.5 and later and only want to monitor your
application locally, you do not need to do anything else. It will be automatically
discovered by JRockit Mission Control.
– If you want to enable your application for remote monitoring, you need to add the
-Xmanagement option to the command line. SSL and authentication will be enabled by

3. Create JRA Reco rding
default. If you do not wish to set up certificates ssl and authentication can be disabled
by providing ssl=false and authenticate=false. Also, if you want to use the
remote discovery feature of the JRockit JVM, you can enable it by setting
autodiscovery=true, for example,
-Xmanagement:ssl=false,authenticate=false,autodiscovery=true
2. Start your Java application and make sure it is running with a load. This way you get the best
possible data collected for the JRA recording.
3. Create JRA Recording

It is simple to create a JRA recording, you only need to select a profile that you want to use. You
can create recordings that either contain all “regular” JRA data or more extensive ones that also
contain thread latency data. This section explains the difference between using the profile “JRA
Recording Normal” without the advanced settings and with the advanced settings. There are
alternative ways to start a JRA recording (see Alternative Ways to Start a JRA Recording).
Note: If you are running JRockit Mission Control on a Windows system, you need to be a
member of the Administrators or the Performance Logs user groups to create a JRA
recording. The typical error message, for not being part of either of these groups, can look
like this:
[perf ] Failed to init virtual size counter:
The instructions for how to use the JRA Recording Normal profile is described in:
z To use the normal recording profile
z To use the normal recording profile with advanced options

For instructions on how to create a JRA recordings that includes latency data, see Creating a JRA
Recording with Latency Data.
To use the normal recording profile

1. Make sure that your application is running and is under load.
If you run the application without load, the data captured from that application will not
show where there is room for improvement.
2. In the JVM Browser, select the JRockit JVM instance you just started or select an entire
folder with running the JRockit JVM instances.
3. Click the Start JRA recording button.

The JRA Recording dialog box appears (Figure 2-2).

Figure 2-2 Start JRA Recording Dialog Box
4. Select the connection you want to record.
5. From the Select recording file drop-down list, choose JRA Recording Normal.
This option is the “classic” JRA recording. This file contains all information that you as a
returning user have found in JRA recordings for previous versions of JRockit Mission
Control and the JRockit JVM.
6. If you are using the JRockit Mission Control Client as an Eclipse plug-in and you want to use
a different filename for the recording file or save it to a different project, click Browse.
The Browse for Project dialog box appears (Figure 2-3).

Note: If you are using a standalone version of the JRockit Mission Control Client or you
don’t want to change the recording filename or project, skip this step and proceed to
step 10.
Figure 2-3 Browse for Project Dialog Box
7. Select the project and folder wherein you want to save the recording.
8. If you want, enter a new Filename.
9. Click Finish.
The dialog box closes.
10. Set a recording time for the duration of the recording in the Recording time field.
11. Select the time unit you wish to use for specifying the recording time (minutes or seconds).
Note: If you set a time that is too short, e.g. shorter than 30 seconds, you will probably not
get enough sample data for the recording to be meaningful.

12. Click Start.

The JRA recording progress window appears. When the recording is finished, it loads in
the JRA tool.
To use the normal recording profile with advanced options

folder with running the JVM instances (if you select a folder, you need to select the JVM you
want to monitor within the JRA Recording dialog box).
4. From the Select recording file drop-down list, choose JRA Recording Normal.
5. Click Show Advanced Options.

A panel with all options for creating a recording become visible (Figure 2-4).
Figure 2-4 JRA Recording with advanced option selected

6. If you are using the JRockit Mission Control Client as an Eclipse plug-in and you want to use
a different filename for the recording file or save it to a different project, click Browse.
The Browse for Project dialog box appears (Figure 2-3).
Note: If you are using a standalone version of the JRockit Mission Control Client or you
don’t want to change the recording filename or project, skip this step and proceed to
step 10.
7. Select the project and folder wherein you want to save the recording.
8. If you want, enter a new Filename.
9. Click Finish.
The dialog box closes.
12. Set a delay for when the recording should start in Delay before starting recording.
It is good to set a delay if you know that your application has, for example, a long
warm-up period.
13. Select none, one, or all of the following options:

– Enable method sampling—records samples of methods.
– Enable GC sampling—records garbage collection events.
– Enable native sampling—records samples of native code.
– Stacktraces—records method stack traces.
– Trace depth—decides how “deep” (how many levels that they contain) the stack traces
should go.
– Sample time—decides how often samples should to be taken. If you set a small
number, samples will be taken more frequently.
– Hardware sampling—records sample data in the same manner as the command-line
option -xxhpm.
– Heap statistics—forces a garbage collection in the beginning and end of the recording
to get Java heap data.

– Thread dumps—creates thread dump data in the beginning and end of recording. If
you set the Thread dump interval, you will also get thread dump data during the
recording.
– Thread dump interval—the time interval for how often thread dumps should be
created. The thread dumps are displayed on the Threads tab in JRA.
– Enable Latency Recording—creates latency data, see Creating a JRA Recording with
Latency Data for more information.
– Latency threshold—sets latency threshold, see Creating a JRA Recording with
Latency Data for more information.
– Enable CPU sampling—records JVM system and user load data and CPU usage. The
information that is recorded is visible on the latency tabs.
– CPU sample interval—sets the time interval for how often CPU sampling should be
performed.
14. Click Start.

the JRA tool.
See Also
There are alternate ways to start a JRA recording, see Starting a Recording with jrcmd and
Starting a Recording From the Command Line.
About JRA Overhead when Recording

The overhead while recording is very low—typically less than two percent. If you enable Java
Lock profiling with the system property –Djrockit.lockprofiling=true the overhead could
be considerable greater, typically between 3% and 25% depending upon the application.
However, since JRA is forcing a full garbage collection at the beginning and at the end of the
recording to generate the heap histogram data, there may be a spike at the beginning and at the
end of a recording. This can be fixed by turning off the option Heap Statistics in the JRA
recording window (see Figure 2-4).
4. View Your JRA Recording in JRA

The recording results in an JRA file that opens automatically in the JRockit Mission Control
Client upon completion. For the JRockit JVM 1.4 versions, the recording file is not transferred

5 . P e r fo r m C h an ges i n A pp l i c a ti o n o r U s e O th e r C o m m an d- l i ne O pt i o n s f o r t he J Ro c k i t J V M
back to the JRockit Mission Control Client, it is saved to the disk where the JRockit JVM is
running and you need to open the file manually by using File > Open.
This topic describes two ways to open a JRA recording: It shows you how:
z To open a JRA recording by dragging and dropping
z To open a JRA recording within the JRockit Mission Control Client
To open a JRA recording by dragging and dropping

1. Locate the JRA recording on your file system.
2. Drag and drop the file into the JRockit Mission Control Client.
To open a JRA recording within the JRockit Mission Control Client

1. In the JRockit Mission Control Client, click File > Open File.
2. Locate and select the recorded file and click Open.
3. Click OK.
The JRA General tab now opens and you can view the data in the recording (see
Figure 6-1).
Note: If you have opened a recording that has been recorded with an older version of JRA, some
fields might not have any relevant data, since that data was impossible to obtain. That
data will appear as “N/A”.
5. Perform Changes in Application or Use Other

Command-line Options for the JRockit JVM
For your second recording, you should make changes to either your Java application or the
command-line options of the JRockit JVM. Typical changes can be setting a different heap size
on the nursery or changing the garbage collector in the JVM. Another good comparison could be
to start your application with a newer version of the JVM, to see if the out-of-the box performance
gives you better and more desired results.

6. Create a New JRA Recording

Create a new recording with the new settings or other JRockit JVM version. The recording that
you are comparing should be of the same length for optimal comparable data. See 3. Create JRA
Recording for information on how to start a recording.
7. Compare and Contrast Two Recordings in JRA

JRA is excellent to use for comparing and contrasting recordings. Open both recordings in JRA
and lay them next to each other to compare the results.
To compare and contrast JRA recordings

1. Create two recordings, one for each setting you wish to try.
2. Open both recordings and lay them out in JRA next to each other (Figure 2-5).
Figure 2-5 Comparing two JRA recordings in JRA

7. Co mpa re an d Cont r ast T wo Re cor din gs in J RA
Figure 2-5 shows the difference in Java heap content between two JRA recordings. The
upper recording has much more dark matter than the lower one. The dark matter can cause
disk fragmentation and will eventually slow down your application.


CHAPTER 3
Alternative Ways to Start a JRA

Recording
The default behavior is to start the JRA recording from within the JRockit Mission Control Client
(see 3. Create JRA Recording), but there are two alternate ways to start a recording. This section
describes the two alternative ways to start a JRA recording.
z Starting a Recording with jrcmd
z Starting a Recording From the Command Line
Starting a Recording with jrcmd

If you run the application without stress, the data captured from that application will not
show where there is room for improvements.
2. Use one of the following commands to initiate a recording:

Windows platforms:
bin\jrcmd.exe <pid> jrarecording time=<jrarecording time>
filename=<filename>
Unix platforms:
bin/jrcmd <pid> jrarecording time=<jrarecording time> filename=<filename>
Where the arguments are:

Al ternative Ways to Start a JRA Recording
– jrarecording time—the duration of the recording in seconds (a good length is 300

seconds, i.e., five minutes).
– filename—the name of the file you want to save the recording to (for example
jrarecording.xml.zip). The file will be created in the current directory of the
JRockit JVM process. It will be overwritten if it already exists.
For example:
bin\jrcmd.exe <pid> jrarecording time=300 filename=c:\temp\jra.xml.zip
Starts a JRA recording of 300s and stores the result in the specified file.
After the recording is initiated, the JRockit JVM prints a message indicating that the
recording has started. When the recording is done, it will print another message; it is now
safe to shut down your application.
Starting a Recording From the Command Line

Use the -XXjra command in combination with an option listed in Table 3-1, for example,
-XXjra:recordingtime to specify the duration of the recording.
Table 3-1 Command Line Startup Options

Option Description
delay Amount of time, in seconds, to wait before recording starts.
recordingtime Duration, in seconds, for the recording. This is an optional parameter. If you
don’t use it, the default is 60 seconds.
filename The name of recording file. This is an optional parameter. If you don’t use it,
the default is jrarecording.xml.
sampletime The time, in milliseconds, between samples. Do not use this parameter unless
you are familiar with how it works. This is an optional parameter.
nativesamples Displays method samples in native code; that is, you will see the names of
functions written in C-code. This is an optional parameter.
methodtraces You can set this to false to disable the stack trace collection that otherwise
happens for each sample. The default value is true.

Starti ng a Re cordi ng From the Command Line
Table 3-1 Command Line Startup Options

Option Description
tracedepth Sets the number of frames that will be captured when collecting stack traces.
Possible value are 0 through 16. The default value is 16.
heapstat=<true | Allows you to enable or disable the tracking of heap statistics.

false> • -XXjra:heapstat=true enables heap statistic tracking
• -XXjra:heapstat=false disables heap statistic tracking.
This tracking is enabled by default but, under certain circumstances can
adversely affect transaction latency. In those situations, it is strongly
recommended that you disable heap statistic tracking.
Note: Setting methodtraces to false can still result in some stack traces being captured. These
stack traces are captured as part of the JRockit JVM’s dynamic optimizations and will
have a depth of 3. If optimizations are turned off (-Xnoopt) these traces will not be
captured.
The startup options that you have used are shown in the VM Arguments tab on the General tab.
See View VM Arguments.
Listing 3-1 shows an example of how you can setup a JRA recording.
Listing 3-1 An example of using the -XXjra startup command:
-XXjra:delay=10,recordingtime=100,filename=jrarecording2.xml
would result in a recording that:
z Commenced ten seconds after the JRockit JVM started (delay=10).
z Lasted 100 seconds (recordingtime=100).
z Was written to a file called jrarecording2.xml (filename=jrarecording2.xml).

Al ternative Ways to Start a JRA Recording

CHAPTER 4
Using the JRockit Runtime Analyzer
z Getting Started with the JRockit Runtime Analyzer
z General Information in a JRA Recording
z Methods and Call Trace Information
z Garbage Collection Events Information
z General Garbage Collector Information
z Java Heap Content Information
z Objects Information
z Code Optimization Information
z Lock Profiling Information
z Start and End Processes Information
z Threads Information
z Using the Latency Tabs
z Latency Log Information
z Latency Graph Information
z Latency Traces Information

Using the JRocki t Runtime Anal yze r
z Adding Comments and Notes to a Recording

CHAPTER 5
Getting Started with the JRockit

Runtime Analyzer
A JRA recording comes with a wealth of information that might seem cumbersome to interpret
at first. You need to keep in mind, however, that the recording should be used when you know
that you have a problem with your application, then the JRA information can help you visualize
those problems so that you have a better chance of fixing them.
This topic gives an overview of the JRA components and how to customize the tool itself. It
includes the following sections:
z Starting JRA
z JRA Overview
z Customizing Your JRA
Starting JRA
There are two ways JRA is started: either automatically when you have created a recording (see
To use the normal recording profile) or when you open an already existing recording (see To open
a JRA recording within the JRockit Mission Control Client).
Note: If you are running a JRockit JVM based on Java 1.4, the JRA tool does not open
automatically when the recording is completed.
JRA Overview
JRA is a multi-tabbed interface, each tab allowing you to monitor different aspects of a JRA
recording. New for Oracle JRockit Mission Control 3.0, is that you can view thread information

G e t ti n g St ar t e d w i t h th e J R o c k i t R un ti m e A na l y z e r
and thread latency information. When all types of recording data has been collected and when all
tabs are activated, the JRA tool includes eleven tabs within the main window (Figure 5-1). When
you view Latency information, the extra tabs (outside the main JRA window) Event Types and
Properties are also used (marked 2 in Figure 5-1).
Note: The number of tabs that are displayed depends on the JRA recording itself (if all sample
data has been collected or not) and settings in the Properties window (see Turning on/off
Tabs).
Figure 5-1 JRA overview
The main JRA window is divided into the following sections:
1. The main JRA window—the available tabs depends on settings in the Preferences window
and the type of data collected in the JRA recording.
2. Tabs that are valid for Latency trouble shooting only.
3. Tabs for different aspects of the JRA recording.

J RA T ab s a t a Gl an ce
JRA Tabs at a Glance

The following information about the tabs are available:
z Getting Familiar with the General Tab
z Getting Familiar with the Methods Tab
z Getting Familiar with the GC General Tab
z Getting Familiar with the GCs Tab
z Getting Familiar with the Heap Tab
z Getting Familiar with the Objects Tab
z Getting Familiar with the Optimizations Tab
z Getting Familiar with the Locks Tab
z Getting Familiar with the Processes Tab
z Getting Familiar with the Threads Tab
z Getting Familiar with the Latency Log Tab
z Getting Familiar with the Latency Graph Tab
z Getting Familiar with the Latency Traces Tab
z Adding Comments and Notes to a Recording
Customizing Your JRA

You can customize your JRA in the following ways:
z Turning on/off Tabs
z Changing Table Settings
z Filtering Information
z Collapsing and Expanding an Information Panel
z Changing Layout of a Tab

Turning on/off Tabs

When you create a JRA recording, there are several options that you can choose to record or not
(see 3. Create JRA Recording). If you decide to exclude something from the recording, the JRA
tool automatically excludes the tab that does not contain any information. This way, you will not
get so many tabs to maneuver within the JRA tool. You can, however, have JRA show all tabs,
by turning on that function in the Preferences window.
To set preferences for the JRA tool

1. Click Window > Preferences.
The Preferences window opens (Figure 5-2).
Figure 5-2 Setting preferences in the JRA tool
2. Select none, one, or both of the JRA preferences:

– Show every tab...—when you choose to see every tab, JRA shows all tabs in the
interface regardless of if the tab contains any information.
– Enable extra information...—the extra information is only useful to support personnel
and this option is used if you have been asked to send a JRA recording to your support
representative.
3. Click Apply for the settings to take effect.
4. Click OK to close the Preferences window.

C ha ng i n g T ab l e S e t t i ng s
Changing Table Settings

JRA lists a lot of information in different tables. These tables can be customized to display
information of your choice. You can also preset the width of the columns in the tables.
Note: You need to change the settings per table, i.e. there is no global change to all tables since
they contain different types of information depending on the tab you are looking at.
To change the settings of the table

1. Click the Table settings button (Figure 5-3).
Figure 5-3 Table settings button
A Table settings window appears (Figure 5-4).

Figure 5-4 Table settings window
2. Select what you want displayed in the table.
3. Set the Min. width and Weight of the column (optional) to a pixel value of your choice.
4. Select Initial sort order for a table item that you want the table to be sorted by.
5. Click OK.
Filtering Information
Some of the information tables can contain lengths of data that can be hard to scroll through.
Instead of scrolling through the long tables, you can filter for the information that you are
interested in viewing.
To filter information
1. Select a table column name for which you want to filter the information. In this example,
Figure 5-5, Pause Time was selected.
2. Enter a number or text for the information you want to see. In this example, Figure 5-5, 60*
was used to see all Pause Times that contains a value starting with 60.

C o l l ap s i n g a nd E x pa nd i ng a n I nf o r m at i o n P an e l
Figure 5-5 Filtering information
Collapsing and Expanding an Information Panel

Sometimes the information on a tab can be cumbersome to work with, then it is good to collapse
the view of the panels that you are not working with or viewing.
To collapse/expand a view
z Click on the small arrow next to a description field (see highlight in Figure 5-6) to collapse
the view of the General Information field.
Figure 5-6 Collapsing a view
Changing to view less values by right clicking a field. The next time you start JRA, you will not
see the specific field.
Changing Layout of a Tab

Sometimes the method names are hard to view in the default horizontal layout, therefore, you
might want to change the layout to a vertical view instead.

To change the layout of a tab

z Click either the Horizontal layout or the Vertical layout button in the right hand corner of
the tab that you are viewing (Figure 5-7).
Note: Not all tabs have this functionality.
Figure 5-7 Horizontal and Vertical layout buttons

CHAPTER 6
General Information in a JRA Recording
The JRA recording contains a lot of data about the application’s behavior, information about the
Oracle JRockit JVM itself, such as the version and which commands were used at the startup of
the JVM. That general information is displayed on the General tab in JRA.
For recordings that have been generated with a JRockit JVM that is older than R26.4, you should
still be able to open them in this version of JRA; however, some fields may be blank, since older
versions of the JRockit JVM did not have the same recording capabilities as newer releases.
Note: Only text fields that require extra explanations have been covered in this documentation.
z Getting Familiar with the General Tab
z Viewing General Information
z Viewing Memory Usage Information
z Viewing Miscellaneous Information
z Viewing Memory Allocation Information
z Viewing Threads Information
z Viewing Exceptions Information

Gene ral I nformation in a JRA Rec ording
Getting Familiar with the General Tab

The General tab (Figure 6-1) contains information on both the JRockit JVM, your system, and
your application.
Figure 6-1 The General tab
The General tab is divided into the following sections:
1. General Information—contains all general information about the JVM, operating system,
recording time, etc.
2. Memory Usage—contains information on how the JRockit JVM is using the memory.

View ing General Informatio n
3. Miscellaneous—contains additional information about a recording. This section is divided

into two tabbed panels:
– VM Arguments—lists all startup options that were used.
– Recording Parameters—lists all the configurable options used during the recording
and the values see for them.
4. Allocation—contains information on how your application allocates memory on the Java

heap.
5. Threads—contains information on thread usage.
6. Exceptions—contains exceptions related information.
Viewing General Information

This panel displays (Figure 6-2) information about the the JRockit JVM version, the operating
system version, number of CPUs that has been used during the recording, etc.
z The value Actual recording time can differ from expected recording time, e.g. if the
application running on the JRockit JVM finished while a recording was still in progress.
z The Maximum heap size is set with a JRockit JVM command-line option.
z The VM information can be information regarding the garbage collection that has been
used.
z The value Number of codeblocks is a JVM internal value. All generated code is divided
into (non-heap) memory blocks called code blocks.

Figure 6-2 General Information panel
Viewing Memory Usage Information

This panel (Figure 6-3) shows a snapshot of the memory usage before and after the recording.
z The value Committed java heap was the current total heap size at the beginning and the
end of the recording. It is less than or equal to the maximum heap size.
Figure 6-3 Memory Usage panel
Viewing Miscellaneous Information

The Miscellaneous panel is a tabbed interface that shows information that can help you better
understand a recording. In this section, you can:
z View VM Arguments
z View Recording Parameters

Viewi ng Mi sce llaneo us Informatio n
View VM Arguments
This panel displays (Figure 6-4) the different command-line options that were used when starting
the JRockit JVM. The options that have been used in the example are the following:
z The JRA recording records latencies (XXjra:latency) has been set (100 seconds).
z The name of the recorded file has been set (filename) and the duration of the recording
(recordingtime).
z The initial, minimum and maximum Java heap has been set (-Xms and -Xmx)
z Some non-standard (Oracle internal) D-options have also been set in this example.
There are many more command-line options that can be set. For comprehensive information on
the different command-line options, please see the Oracle JRockit JVM Command-Line
Reference.
Figure 6-4 VM Arguments
View Recording Parameters

This panel displays all configurable options used in the current recording and the values attributed
to those options.
Figure 6-5 Recording Parameters
You can determine which options you can see by specifying a filter. To filter options, do the
following

1. Select a table column name for which you want to filter the information. In Figure 6-6, the
Key column was selected.
2. Enter text for the information you want to see; for example, in Figure 6-6, cpu* was entered
to limit the options displayed to just those beginning with the text string “cpu”.
Figure 6-6 Filtered Recording Parameters
Viewing Memory Allocation Information

This panel displays information about how the JRockit JVM is allocating memory on the Java
heap (Figure 6-7). A Thread Local Area (TLA) is a JRockit JVM internal value. It is a small
memory area, local to a thread, where the JVM can allocate small objects without having to take
the heap lock. For an in-depth explanation of how TLA works, please see Setting the Thread
Local Area Size in the Oracle JRockit JVM Diagnostics Guide. See also -XXtlaSize in the
Oracle JRockit JVM Command-Line Reference for more information on how to set different
values of the TLA size.
z Preferred thread local area (TLA) size is the value that you have set with the
command-line option -XXtlasize:preferred <size>.
z Minimum thread local area size is the value that you have set with the command-line
option -XXtlasize:min <size>.
z Ratio of bytes for large/small objects. Per default, the JRockit JVM considers an object
to be large if it is larger than the thread local area size; it is small if it would normally fit in
a thread local area. Large objects are always allocated in the old space (second generation)
of the heap, never in the nursery.
z The Number (#) free list misses is a JRockit JVM internal value. The JRockit JVM has a
list of free memory blocks on the Java heap. During allocation, an object is normally put in
the first free block on the “free list.” If it does not fit there, the JVM will try the next
block, and the next, etc. Each block where the code block did not fit is considered a “free
list miss.”

Vie w i n g T hr e a ds I nf o r m at i o n
Figure 6-7 Allocation panel
Viewing Threads Information

This panel displays (Figure 6-8) information on the number of Java threads that existed both
before and after the recording.
z The value of Total number of threads before/after recording shows how many threads
were active before the recording started and how many were active when the recording
ended.
z The value of Number of deamon threads before/after recording is the number of

deamon threads. A deamon thread is a thread that runs in the background to support the
runtime environment, for example, a garbage collector thread. The JVM exists when all
non-daemon threads have completed.
z The value Number of threads started during recording shows how many threads were
started.
z The value System total of # (number) context switches per second is fetched from the
operating system. An unusually high context switch value compared to other applications
may indicate contention in your application.
Figure 6-8 Threads panel

Viewing Exceptions Information

This panel displays (Figure 6-9) information on the total number of Java exceptions that are
thrown during a recording. This includes both caught and uncaught exceptions. Excessive
exception throwing can be a performance problem. Hardware generated exceptions are
originating from a “trap” in the hardware and are usually the most “expensive” kinds of
exceptions.
Figure 6-9 Exceptions information

CHAPTER 7
Methods and Call Trace Information
Methods where the Oracle JRockit JVM spends most of its time are called hot. Once you have
identified such a method, you might want to investigate it to see if it is a “bottleneck” for the
application or not. The way that the JRockit JVM collects method information is via a sampling
thread that is called the hotspot detector. It uses statistical sampling to find Java methods that are
candidates for optimization. The samples are collected by iterating through the Java threads in the
virtual machine and suspending them one at a time. The current instruction pointer of the
suspended thread is used to lookup in which Java method the thread is currently executing. The
invocation count of the method is incremented and the method is added to a queue of methods to
be optimized if the invocation count exceeds a certain threshold.
The JRA recording system makes use of the hotspot detector by setting it to a high sampling
frequency during the recording and directing the samples to the .jra file.
z Getting Familiar with the Methods Tab
z Viewing Hot Methods
z Viewing Predecessors and Successors
z Jumping to Application Source
Getting Familiar with the Methods Tab

The Methods tab (Figure 7-1) lists the top hot methods, with its predecessors and successors,
that were recorded.

Me thods and Call Trac e Information
Figure 7-1 The Methods tab
The Methods tab is divided into the following sections:
1. Top Hot Methods—a listing of the top hot methods. Click on the different table headings to
get a different sort order.
2. Filter column—see Filtering Information on how to use this function.
3. Predecessors—a listing of all preceding methods to the method that you have selected in the
Top Hot Methods list. If you have selected many methods, there will not be any information
shown in this panel.
4. Successors—a listing of all succeeding methods to the method that you have selected in the
Top Hot Methods list. If you have selected many methods, there will not be any information
shown in this panel.

V i e wi ng H o t M e t ho d s
Viewing Hot Methods

The method sampling in the JRockit JVM is based on CPU sampling. This requires that you put
load on the system to get any samples. The Top Hot Methods table (Figure 7-2) lists all methods
sampled during the recording and sorts them with the most sampled method s first. These are the
methods where most of JVM’s time is spent.
Figure 7-2 Top Hot Methods shown
Note: If your recording has native sampling enabled during the recording, you can see methods
prefixed by jvm, which are native methods in the JRockit JVM.
Use the filtering function to find the method you are looking for, see Filtering Information.
Viewing Predecessors and Successors

By selecting a method in the Top Hot Methods table, you can see its sampled Predecessors and
Successors (Figure 7-3). The predecessors are the methods that call the selected method and the
successors are the methods that the selected method calls.

Figure 7-3 Viewing Predecessors and Successors
The number within brackets of a particular predecessor or successor is the number of sampled
call traces of which the method is part. The percentage shows how common a particular path is
in the method tree. If you see methods that are called a lot from the JRockit JVM, you might want
to investigate if that method is causing your application to run slower than necessary.
Jumping to Application Source

If you are using JRA as an Eclipse plug-in, you can jump from the Top Hot Methods table and
the Predecessors and Successors trees directly to the source code. A feature called
Jump-to-Source allows you not only to see the name of a “problem” method displayed in the GUI,
but lets you jump from the displayed method directly to that method’s source, where you can
evaluate the code to see what might be causing the problem. This feature extremely is useful in
helping you locate and debug coding errors that are creating runtime problems for your
application.
To jump to the source code

1. In the Top Hot Methods table and the Predecessors and Successors trees, right-click the
problem method or class to open a context menu.

Ju mpin g to Ap pl ica ti on Sou r ce
2. Select Open Method.
3. The source code appears in a separate editor.


CHAPTER 8
General Garbage Collector Information
The GC General tab shows an overview of information about all garbage collections (GC) that
took place during the recording. The information includes, amongst other, the total number of
pause times and when and how the garbage collector has changed strategy.
z Getting Familiar with the GC General Tab
z Viewing General Garbage Collection Information
z Viewing Garbage Collection Call Tree Information
z Viewing Garbage Collection Strategy Changes Information
Getting Familiar with the GC General Tab

The GC General tab (Figure 8-1) shows general information about a garbage collection, its call
tree, and what garbage collection strategies have taken place.

Gene ral Garbage Co llec to r Information
Figure 8-1 The GC General tab
The GC General tab is divided into the following sections:
1. General—this panel shows overall statistics about the garbage collections during the entire
JRA recording.
2. Garbage Collection Call Tree—this panel is a collection of all call traces that were sampled
for all garbage collections for the JRA recording.
3. GC Strategy Changes—this table lists when a garbage collection strategy took place and
how it changed.
Viewing General Garbage Collection Information

The General panel (Figure 8-2) shows general garbage collection information such as the total
number of garbage collections during the recording and the duration of all pause times due to

Vie wing Garbage Colle cti on Call Tree Informatio n
garbage collection. You can use this information to, for example, see whether your application is
coming down to desired pause time averages or not.
Figure 8-2 General Garbage Collection Information
Viewing Garbage Collection Call Tree Information

The Garbage Collection Call Tree panel (Figure 8-3) shows all call traces during the recording
that triggered a garbage collection. The number within the brackets (next to the garbage bin icon)
is the total number of garbage collection rounds that were performed during the JRA recording.
Expand the call tree to see in which methods the garbage collection has taken place.
Figure 8-3 Garbage Collection Call Tree Information
Viewing Garbage Collection Strategy Changes

Information
The Garbage Collection Strategy Changes table (Figure 8-4) lists when the garbage collector
has changed strategy, for example, the JRockit JVM has been set to run for best throughput
(-Xgcprio:throughput, GC Prio in Figure 8-4), then the JRockit JVM changes strategy in
runtime to best reach this goal. The strategy change can, for example, be from singleParPar to
genParPar. The strategy changes are listed under New Strategy. The old strategies are listed
under Generational, Mark Phase, and Sweep Phase.

Gene ral Garbage Co llec to r Information
Note: These strategy changes only happen if you are running the JRockit JVM with the default
garbage collector option, -Xgcprio.
Figure 8-4 Garbage Collection Strategy Changes Information
In the example seen in Figure 8-4, there has been one strategy change for the garbage collector.
Use the filtering function to find a specific garbage collection, see Filtering Information.

If you are using JRA as an Eclipse plug-in, you can jump from the Garbage Collection Call tree
directly to the source code. A feature called Jump-to-Source allows you not only to see the name
of a “problem” method displayed in the Garbage Collection Call tree, but lets you jump from
the displayed method name directly to that method’s source, where you can evaluate the code to
see what might be causing the problem. This feature extremely is useful in helping you locate and
debug coding errors that are creating runtime problems for your application.

1. In the Garbage Collection Call tree, right-click the problem method to open a context menu.

CHAPTER 9
Garbage Collection Events Information
The GCs tab shows detailed information about each garbage collection (GC) event that has
occurred. The tab contains a graph for Java heap usage before and after each garbage collection
as well as detailed garbage collection information for each collection.
z Getting Familiar with the GCs Tab
z Changing Focus on GC Chart
z Viewing Specifics about Garbage Collections
z Viewing the Detailed Information About the Garbage Collection
z Viewing Information on the Common Garbage Collection Tab
z Viewing Information on the GC Method Call Tree Tab
z Viewing Information on the Old/Young Collection Tab
z Viewing Information on the Cache Lists Tab
z The Pause Time Tab

Garbage Coll ection Events Informatio n
Getting Familiar with the GCs Tab

The GCs tab visualizes how and when a garbage collection has occurred during the running of
the application (Figure 9-1). It also shows specific information for each garbage collection.
Figure 9-1 The GCs tab
The GCs tab is divided into the following sections:
1. GCs Overview timeline—this timeline shows the entire recording in its full length (when you
initially open your recording). You can use this to refocus the Heap Usage graph, see
Changing Focus on GC Chart.
2. Heap Usage graph—this graph shows heap usage compared to pause times and how that
varies during the recording. If you have selected a specific area in the GC Chart, you will only
see that section of the recording. You can change the graph content in the Heap Usage

Ch an gi ng Foc us on GC Cha rt
drop-down list (marked 6 in Figure 9-1) to get a graphical view of the references and
finalizers after each old collection.
3. Garbage Collections events—this list shows all garbage collection events that have taken
place during the recording. When you click on a specific event, you will see a corresponding
flag in the Heap Usage graph for that particular event, see Viewing Specifics about Garbage
Collections.
4. Details—this panel contains all the details about the specific garbage collection round. When
you select a garbage collection in the Garbage Collection list, the tabs in the Details panel
changes depending on if you have selected an old collection or a young collection.
5. Chart Configuration—this section allows you to change the appearance on the active chart.
6. Drop-down list and Show strategy changes—the drop-down list allows you to select one of
these views on the Heap Usage chart:
– Heap Usage
– References and finalizers
– Pause Time
– CPU
If you select Show strategy changes, you will see when the JRockit JVM has changed
garbage collection strategy.
7. Move and Zoom buttons—these buttons are used with the GCs Timeline.
Changing Focus on GC Chart

Depending on how long your JRA recording is, the GC Chart can be quite cumbersome to view
in full mode; therefore, you can refocus the chart. by dragging the handles on the slide bar to the
section of the recording that you want to view. Once you have set the side on the slide bar, you
can slide that section to the position of the chart that you are interested in studying.
The two ways to refocus on the GC Chart are described here:
z To change focus on the Heap Usage chart
z To use the Move and Zoom buttons for the GC Chart
To change focus on the Heap Usage chart

1. Click and drag the handles on both sides on the GC Chart (Figure 9-2).

Figure 9-2 The GC Chart zoom function
2. Drag the GC Chart into the desired position (Figure 9-3).
Figure 9-3 The GC Chart
To use the Move and Zoom buttons for the GC Chart

1. Click the Move forward or Move backward buttons (marked 1 in Figure 9-4) to first
decrease the GC Chart view.
Figure 9-4 Move and Zoom buttons
2. Click either of the Move buttons to slide the focus on the GC Chart.
3. Click the Zoom in or Zoom out buttons (marked 2 in Figure 9-4) to decrease or increase the
visible span of the GC Chart.
Viewing Specifics about Garbage Collections

The Garbage Collections table on the GCs tab is a list of all garbage collections that have taken
place during the recording. It lists all garbage collection events during the recording, provided
that the garbage collection sampling was enabled. A garbage collection can be an old collection,
which is a garbage collection in the old space of the Java heap or a young collection, which is a

Vi ewing Specific s about Garbage Coll ections
garbage collection in the young space (nursery). For more information on garbage collections,
please see Garbage Collection in Oracle JRockit in the Oracle JRockit JVM Diagnostics Guide.
z To view one garbage collection in the GC Chart
z To view many garbage collections in GC Chart
To view one garbage collection in the GC Chart

1. Scroll in the Garbage Collection list to the garbage collection you want to view.
2. Click on that garbage collection.

The garbage collection index number is now visible in the GC Chart and the Details panel
has also changed to show all the specifics about that garbage collection.
The Details panel changes name depending on if the selected event is an old collection or a
young collection (Figure 9-5).
Figure 9-5 Viewing one garbage collection
To view many garbage collections in GC Chart

1. Scroll the Garbage Collections list.
2. Click and hold either the Shift key or Ctrl key to select multiple collections.

The garbage collection index numbers are now visible in the GC Chart (Figure 9-6).
Note: The garbage collection event that was last selected is the one that is displayed in the
Details panel.
Figure 9-6 Viewing multiple garbage collections
Viewing the Detailed Information About the Garbage

Collection
When you select a garbage collection, the Details panel of the GCs tab changes name to either
Details - Old Collection or Details - Young Collection depending on the type of garbage
collection you have selected. You will also see different sets of tabs that contain specific
information about the garbage collection that you have selected (Figure 9-7).

V i e w i ng I n f o rma t i o n o n th e Co m m o n Ga r ba ge C o l l e c ti o n T a b
Figure 9-7 Tab differences when viewing old and young collections
Each one of these tabs are described here. As much of the information in the tabs are fairly
self-explanatory, those types of details will not be covered in the documentation.
This section describes the following tabs:
z Viewing Information on the Common Garbage Collection Tab
z Viewing Information on the GC Method Call Tree Tab
z Viewing Information on the Old/Young Collection Tab
z Viewing Information on the Cache Lists Tab
z The Pause Time Tab
Viewing Information on the Common Garbage Collection

Tab
The Common tab (Figure 9-8) displays information such as start time and end time of the
garbage collection.

Figure 9-8 The Common garbage collection tab
z Sum of Pauses—the sum of all pause times in milliseconds that the garbage collector stops
all threads in the JRockit JVM. This is not the same as end time-start time in the case of a
concurrent garbage collector.
z Start/End Time—the times when the garbage collection started and ended, counted in
milliseconds from when the JRockit JVM started.
z Heap Usage Before/After—the used heap size before or after the garbage collection.
z Committed Heap Size—the total size of the heap (used plus unused memory) after the
garbage collection.
z Size of Promoted Objects (and number of Promoted Objects)—the size (and the amount)
of the objects that have been promoted to the old space.
z References—there are several types of references collected during a recording. For

information on what a reference is, see Viewing Reference Objects in the Diagnostics
Guide.
z Finalizer Queue Length (and Before)—the finalizer queue length.
z Generation—Indicates whether the garbage collector performed an old or young collection

(see Generational Garbage Collection in the Diagnostics Guide for more information on
generational garbage collection).

Vie win g I nf or mat ion on t he GC Me thod Cal l Tree Tab
Viewing Information on the GC Method Call Tree Tab

The GC Method Call Tree tab (Figure 9-9) shows an aggregation of the call traces of the threads
triggering a garbage collection.
Figure 9-9 The GC Method Call Tree tab
Viewing Information on the Old/Young Collection Tab

The name of this tab is dynamically changed when you select a garbage collection instance in the
Garbage Collections table. Here you find information about nursery, mark and sweep pause
times, etc. (Figure 9-10).
Figure 9-10 The Old/Young Collection tab
z Nursery Size Before/After—indicates the free space in the nursery before and the free
space in the nursery after the garbage collection (in some cases the nursery size increases).
The information below is only valid for old collections:

z Nursery Start/End Position—the starting and ending position in the memory address of
nursery.
z Mark/Sweep Phase Time—the time spent in the marking and sweep phases, measured in
milliseconds.
z Compacted Size—the size of the heap that has been compacted in the garbage collection.
z Compaction Ratio—the ratio of heap size before and after the compaction, measured in
percent.
z Desired/Actual evacuation—the desired evacuation is the size of the area on the Java
heap that you want to evacuate and the actual evacuation is the size of the area that the
JRockit JVM managed to evacuate. The value for actual evacuation can be smaller than the
desired due to temporarily pinned objects (objects that are not allowed to be moved during
garbage collection). The evacuation takes place during compaction or shrinking of the Java
heap.
z GC Reason—indicates the reason for doing this garbage collection.
Viewing Information on the Cache Lists Tab

The Cache Lists tab (Figure 9-11) displays the specification for the different cache lists. Each
cache list contains settings for upper and lower cache size.
Figure 9-11 The Cache Lists tab
z Index—this is the identification number for the cache list.
z #free blocks—the number of free blocks in the cache list.
z Cache size—the total size of this cache list.
z Avg free block size—the average size of each free memory block in the cache list.
z Low limit—the lower limit of a free memory block. There will be no smaller memory
block than this in the selected cache list.

The Pause Time Tab
z High limit—the upper limit of a free memory block. There will be no larger memory
blocks than this in the selected cache list.
The Pause Time Tab

The information under the Pause Time tab is mainly intended for the JRockit team’s internal use
when you have sent a JRA recording for analysis to the JRockit engineering team.
z GC Pause—this column displays the names of the pauses (the main entry in the tree
structure). If you are running a parallel garbage collector, then there will only be one pause
per garbage collection. For the concurrent garbage collector, there can be several pauses
during one garbage collection. The pauses consists of pause parts that can help the JRockit
engineering staff to analyze why certain pauses are longer than others.
Note: During a pause, the application is standing still.
z Duration—this is the length, measured in milliseconds, of the pause.
z Start/End—this is the start and end time, measured in milliseconds. You can change how
the time is displayed by right-clicking in the table and select Start and then the value for
the time.

If you are using JRA as an Eclipse plug-in, you can jump from the GC Method Call tree directly
to the source code. A feature called Jump-to-Source allows you not only to see the name of a
“problem” method displayed in the GUI, but lets you jump from the displayed method name

directly to that method’s source, where you can evaluate the code to see what might be causing
the problem. This feature extremely is useful in helping you locate and debug coding errors that
are creating runtime problems for your application.

1. In the GC Method Call tree, right-click the problem method to open a context menu.

C H A P T E R 10
Java Heap Content Information
The Heap tab gives a quick overview of what the memory in the Java heap consists of in you
application. The overview displays how the heap looked at the end of the recording and it also
shows compiled information about the status of the heap during the entire recording.
This section contains the following topics:
z Getting Familiar with the Heap Tab
z Viewing the Heap Snapshot at the End of the Recording Information
z Viewing the Heap Contents Information
z Viewing the Free Memory Contribution Information
Getting Familiar with the Heap Tab

The Heap tab depicts Java heap contents and free memory distribution (Figure 10-1).

Ja v a H e a p C o nt e nt I nf o r m at i o n
Figure 10-1 The Heap tab
The Heap tab is divided into the following sections:
1. Heap Snapshot at the End of the Recording—this panel contains all the specifics about
your heap at a glance.
2. Heap Contents—this graph gives a visual overview of the distribution of different sizes of
objects. The table below the graph gives the exact data for each category of memory.
3. Free Memory Contribution—this graph gives a visual overview of the distribution of the
different chunks of free memory that there is on the heap. The table below the graph gives the
exact data for each category of memory.
Viewing the Heap Snapshot at the End of the Recording

Information
When JRA stops recording, it calculates the value of the committed heap size, which is how much
heap the application has been allowed to use. This size can be set by the -Xmx flag.

V i e w i n g t he H e a p C o nt e nt s I nf o r m at i o n
The memory that is considered large object chunks, is the total amount of memory on the heap
that the Java application is allowed to use for large objects (64 KB to 512 kB).
The memory for the pinned object chunks is the amount of memory that is occupied by pinned
objects. A pinned object is both referenced by another object in the application and is not allowed
to be moved for compaction purposes, for example, i/o buffers that are accessed from native
methods (native i/o). The number of pinned object chunks shows a value of how many object
that are pinned.
Dark matter is memory that is free, but cannot be used due to the physical layout of the memory
chunk (i.e. it might be too small for the application to allocate). Dark matter can cause
fragmentation on the disk.
Viewing the Heap Contents Information

The Heap Contents pie chart gives a graphic overview of the distribution of objects on the heap.
The color coding helps you determine how much of the heap that consists of large, small, and
pinned object chunks as well as how much memory is considered dark and how much is free. The
amount of dark matter indicates how much space on the Java heap that is wasted due to
fragmentation. It is normal to have a certain amount of dark matter on the heap.
For information on how to minimize dark matter, see Minimize Dark Matter in the Oracle JRockit
JVM Diagnostics Guide.
The table below the pie chart (Figure 10-2) lists all objects with the exact data: memory in MB
and percentage that they occupy of the heap.
Figure 10-2 Heap content table
Viewing the Free Memory Contribution Information

The Free Memory Contribution pie chart gives a graphic overview of how the free memory is
distributed in free blocks of different sizes on the Java heap. The table below the pie chart
(Figure 10-3) lists all block sizes by category.

Ja v a H e a p C o nt e nt I nf o r m at i o n
Figure 10-3 Free memory content table
The block sizes are categorized by the following entities: small, medium, large, and very large.
The block sizes are multiples of the minimum block size set at startup (default 2kB). You set the
minimum block size with the option -XXminblocksize.
Below are the multiples used for the different block sizes:
z Small: 1–4
z Medium: 4–32
z Large: 32–256
z Very large: 256 and up

C H A P T E R 11
Objects Information
The Objects tab displays the most common types and classes occupying the Java heap at the
beginning and at the end of the JRA recording.
z Getting Familiar with the Objects Tab
z Viewing Start of Recording Information
z Viewing End of Recording Information
Getting Familiar with the Objects Tab

At the beginning and end of a recording session, snapshots are taken of the most common types
and classes of object types that occupy the Java heap, that is, the types which instances in total
occupy the most memory. The results are shown on the Object tab (Figure 11-1). Abnormal
results in the object statistics might help you detect the existence of a memory leak in your
application.

Obje cts Information
Figure 11-1 The Objects tab
The Objects tab is divided into the following sections:
1. Start of Recording—this table lists the most common types on the heap at the beginning of
the recording.
3. End of Recording—this table lists the most common types on the heap at the end of the
recording.
Viewing Start of Recording Information

When JRA starts a recording it looks at the Java heap to see which types occupy the most memory
in the used heap space. That information is listed under the Start of Recording table
(Figure 11-2).
Use the filtering function to find the object you are looking for, see Filtering Information.

Vie wing End of Rec ording Informatio n
Figure 11-2 Start of Recording table
Viewing End of Recording Information

Right before JRA stops a recording it looks at the Java heap to see which types occupy the most
memory in the used heap space. That information is listed under the End of Recording table
(Figure 11-3).
Use the filtering function to find the object you are looking for, see Filtering Information.
Figure 11-3 End of Recording table

If you are using JRA as an Eclipse plug-in, you can jump from both the Start of recording and
End of recording tables directly to the source code. A feature called Jump-to-Source allows you
not only to see the name of a “problem” class displayed in the Objects tab, but lets you jump from
the displayed class name directly to that class’s source, where you can evaluate the code to see
what might be causing the problem. This feature extremely is useful in helping you locate and
debug coding errors that are creating runtime problems for your application.

Obje cts Information

1. In the Start of recording or End of recording table, right-click the problem class to open a
context menu.
2. Select Open Type.

C H A P T E R 12
Code Optimization Information
The Oracle JRockit JVM continuously looks for ways to optimize code. The Optimizations tab
displays the methods that were optimized by the adaptive optimization system in the JRockit
JVM during the recording.
z Getting Familiar with the Optimizations Tab
z Viewing Optimization Information
z Viewing Methods Optimized During Recording Information
Getting Familiar with the Optimizations Tab

JRA records all optimization events that occur during the course of the recording. the JRockit
JVM uses JIT compilation for the initial conversion to machine code. The most commonly used
methods are then further optimized during the application run. This information is then displayed
in the Optimizations tab (Figure 12-1).

Code Optimizati on Informatio n
Figure 12-1 Optimizations tab
The Optimizations tab is divided into the following sections:
1. Optimization—this panel displays the before and after scenario of the optimizations that
have taken place.
2. Methods Optimized During Recording—this table lists which methods that have been
optimized during the recording, i.e. this is necessarily not a full list of all optimizations that
are performed for your application.
Viewing Optimization Information

The Optimization panel (Figure 12-2) contains information on how many optimizations have
taken place and the total duration of the optimizations. You can also see how many JIT
compilations have been performed and the time the JRockit JVM took to compile those. For more
information on JIT compilation, see Understanding JIT Compilation and Optimizations.

View ing Methods Opti mized Du r i n g R e c o r di n g I nf o r m at i o n
Figure 12-2 Optimization panel
Viewing Methods Optimized During Recording

Information
The Methods Optimized During Recording table (Figure 12-3) lists all methods that were
optimized during the JRA recording. Here you can study the size changes of each method that has
been optimized.
Note: Some optimizations, such as inlining, causes the method size to increase.
Use the filtering function to find the method you are looking for, see Filtering Information.
Figure 12-3 Methods Optimized During Recording table

If you are using JRA as an Eclipse plug-in, you can jump from the Methods Optimized During
Recording table directly to the source code. A feature called Jump-to-Source allows you not only
to see the name of a “problem” method displayed in the table, but lets you jump from the
displayed method directly to that method’s source, where you can evaluate the code to see what
might be causing the problem. This feature extremely is useful in helping you locate and debug
coding errors that are creating runtime problems for your application.

Code Optimizati on Informatio n

1. In the Methods Optimized During Recording table, right-click the problem method to open
a context menu.

C H A P T E R 13
Lock Profiling Information
The Locks tab shows comprehensive information about lock activity for the application JRA is
monitoring (Java locks) and the Oracle JRockit JVM itself (native locks). You need to enable the
lock profiling data recording capability before you start the profiling of your application. If you
have not enabled the lock profiling data recording, the lock profiling tables are blank on the
Locks tab. For more information on locks, please refer to About Thin, Fat, Recursive, and
Contended Locks in Oracle JRockit.
z Getting Familiar with the Locks Tab
z Java Locks Profiling
z Enabling Java Lock Profiling Data
z Native Lock Profiling
z Enabling Native Locks Information
Getting Familiar with the Locks Tab

The Locks tab displays lock information for both your application and the JRockit JVM
(Figure 13-1).

Loc k Pro fil ing Information
Figure 13-1 Locks tab
The Locks tab is divided into the following sections:
1. Java Locks—this table lists all locks in your application.
3. Native Locks—this table lists all locks in the JRockit JVM.
Java Locks Profiling

The information that is displayed under the Java Locks chart (Figure 13-2) shows the number of
locks of the threads in your application. You see information on the number of fat uncontended
and contended locks, thin uncontended and contended locks, thin and fat recursive locks, and fat
sleeping locks. For more information on locks, please refer to About Thin, Fat, Recursive, and
Contended Locks in the Oracle JRockit JVM.
Use the filtering function to find the Java locks you are looking for, see Filtering Information.

Java Lock s Pro fi ling
Figure 13-2 Java Locks
Enabling Java Lock Profiling Data

To record Java lock profiling data, you need to enable it from the command line when you start
the JRockit JVM. If your the Java Locks table is blank, it is not enabled.
To enable Java lock profiling data

z Issue the command -Djrockit.lockprofiling at the JRockit JVM command line.
For example:
java -Djrockit.lockprofiling=true -XXjra:<AnyJRAParam> -jar MyApplication.jar
Showing Classes Banned from Lazy Unlocking

You can show classes that have been banned from lazy unlocking in the Java Locks table;
however, the Banned Classes column is hidden by default. This column shows that the class/type
has been banned for lazy unlocking”
To unhide the Banned Classes column

z Open the Filter Column drop-down list and select Lazy Banned, as shown in Figure 13-3.
Figure 13-3 Selecting Lazy Banned

Java Lock Profiling Overhead

If you enable Java Lock profiling with the system property –Djrockit.lockprofiling=true
the overhead could be considerable greater, typically between 3% and 25% depending upon the
application.
Native Lock Profiling

If you are looking at a recording of the JRockit JVM 5.0 or later, the recording includes
information about native locks (Figure 13-4). Native locks are locks in the JRockit JVM internal
code and is nothing your application can control.
Use the filtering function to find the Java locks for which you you are looking; see Filtering
Information.
Figure 13-4 Native Locks
If you find high contention on a JRockit JVM internal lock that might be causing issues for your
application. Either contact Oracle support or contact the JRockit support team through the Oracle
JRockit Newsgroup at the dev2dev web site.
Enabling Native Locks Information

Lock profiling data can only be generated from the command line. If you have no information
displayed in the Locks tab, the native sampling was not enabled during the recording. See 3.
Create JRA Recording for information on how to enable native sampling.

If you are using JRA as an Eclipse plug-in, you can jump from the Java Locks table directly to
the source code. A feature called Jump-to-Source allows you not only to see the name of a
“problem” class displayed in the table, but lets you jump from the displayed class name directly

to that class’s source, where you can evaluate the code to see what might be causing the problem.
This feature extremely is useful in helping you locate and debug coding errors that are creating
runtime problems for your application.

1. In the Java Locks table, right-click the problem class to open a context menu.
2. Select Open Type.


C H A P T E R 14
Start and End Processes Information
The Processes tab lists which processes were running during the start and the end of the JRA
recording. The information found on this tab is mostly geared towards engineers within the
customer support team, which uses the information to get a picture of which applications that
were running on the machine when, for example, a crash has occurred. This tab is not visible by
default, so you need to turn it on before you can view that information.
z Turning on the Processes Tab
z Getting Familiar with the Processes Tab
z Snapshot of Processes at Beginning and End of Recording
z Detailed Processes Information
Turning on the Processes Tab

The Process tab is not visible unless you have selected the option Enable extra information for
JRPG CCE in the JRA Preferences
To turn on the Processes tab

The Preferences window appears.
2. Click JRockit Mission Control > Runtime Analyzer (JRA).

Start and End Proces ses Info rmatio n
3. Select the Enable extra information for JRPG CCE option (Figure 14-1).
Figure 14-1 Preferences window
4. Click Apply.
5. Click OK.
If you have a JRA recording open when you change this preference, you need to close it
and then open it again for the Processes tab to become visible.
Getting Familiar with the Processes Tab

The Processes tab displays start and end information of running processes (Figure 14-2).
Note: You need to enable the Processes tab for it to be visible in JRA (see To turn on the
Processes tab).

Snapshot of Proc ess es at Begi nn i ng a nd En d o f R e c o rd i n g
Figure 14-2 Processes tab
The Processes tab is divided into the following sections:
1. Snapshot of the processes running on the machine at the start and at the end of the
recording—this table lists all processes that were active either during the start or the end of
the recording or both.
3. Process—this panel details the processes information.
Snapshot of Processes at Beginning and End of

Recording
The information that is displayed under the Snapshot view (Figure 14-3) lists all processes that
were running at the start of the recording and at the end of the recording.
Use the filtering function to find the process you are looking for, see Filtering Information.

Start and End Proces ses Info rmatio n
Figure 14-3 Snapshot view
Detailed Processes Information

When selecting a process in the Snapshot view, you see a list of all details for that process at the
bottom of the tab (Figure 14-4). The path, the name of the executable, if the process was present
during start and end, the process ID, and also if the process was started with a command-line
option.
Figure 14-4 Detail process view

C H A P T E R 15
Threads Information
The Threads tab lists all thread dumps that have been taken during the recording. If no Thread
Dump interval (in the recording options) is specified, the recording will contain a thread dump
from the start and the end of the recording. A thread dump reveals information about an
application’s thread activity that can help you diagnose problems and better optimize application
and JVM performance; for example, thread dumps automatically show the occurrence of a
deadlock. Deadlocks bring some or all of an application to a complete halt.
The information found on this tab is mostly geared towards engineers within the customer support
team. This tab is not visible by default, so you need to turn it on before you can view that
information.
Note: For comprehensive information on how create and use a thread dump, please see the
Using Thread Dumps section in the Oracle JRockit JVM Diagnostics Guide.
z Turning on the Threads Tab
z Getting Familiar with the Threads Tab
z List of Times when Thread Dump is Taken
z Thread Dump Information
Turning on the Threads Tab

The Threads tab is not visible unless you have selected the option Enable extra information
for JRPG CCE in the JRA Preferences.

Th rea ds In fo rma ti on
To turn on the Threads tab

The Preferences window appears.
2. Click JRockit Mission Control > Runtime Analyzer (JRA).
3. Select the Enable extra information for JRPG CCE option (Figure 15-1).
Figure 15-1 Preferences window
4. Click Apply.
5. Click OK.
If you have a JRA recording open when you change this preference, you need to close it
and then open it again for the Threads tab to become visible.
Getting Familiar with the Threads Tab

The Threads tab lists the available thread dumps and by clicking on a specific time when a thread
dump was created, you see the entire thread dump in the Thread dump (Figure 15-2).
Note: You need to enable the Threads tab for it to be visible in JRA (see To turn on the Threads
tab).

L ist of T im es whe n T hr ea d D u m p i s T ake n
Figure 15-2 Threads tab
The Threads tab is divided into the following sections:
1. Thread dumps taken at various times during the recording—this table lists the times
when a thread dump has been taken.
3. Thread dump—this panel displays the actual content of the selected thread dump.
List of Times when Thread Dump is Taken

The information that is displayed under the Thread dumps taken at various times during the
recording table (Figure 15-3) shows when a thread dump was taken. You set the interval for
taking thread dumps under the advanced option when you create a JRA recording (see To use the
normal recording profile with advanced options).
Use the filtering function to find the specific thread dump, see Filtering Information.

Th rea ds In fo rma ti on
Figure 15-3 Time view
Thread Dump Information

When selecting a thread dump in the Thread dump list view, the entire thread dump is displayed
in the Thread dump panel (Figure 15-4).
Note: To understand the information in the thread dump, please see the Using Thread Dumps
section in the Oracle JRockit JVM Diagnostics Guide.
Figure 15-4 Thread dump output

C H A P T E R 16
Using the Latency Tabs
Finding performance bottlenecks within your Java application is a bit of a detective’s work. You
know what the symptoms of the problem are, for example, the application is running really slow
but the CPU isn’t saturated. Where to start looking for clues to such an issue is tricky since most
profiling tools for Java applications only pinpoint where in the code your application is spending
the most time to run (which is a good start). What these tools tend to miss, however, or not show
at all is where in the application stops and waits occur, i.e. where the application spends time
being idle.
These stops and waits can be caused by poor memory management, such as limited heap space
or a poorly managed heap that requires too many garbage collections. On the other hand the stops
and waits can be latencies caused by multi-threaded applications that spend much of the processor
time waiting, blocking, or sleeping. These problems have previously been hard to detect but now
the JRA system is able to record latencies within your application and visualize running threads
with their events in an easy to understand manner.
This section of the help gives you an overview of how you can use the latency tabs in JRA to work
you way down to a Java application latency. In addition, you get one example of how a Java
application that contains latencies looks on the Latency Graph tab and you will get an example
workflow of how to use all latency tabs together. All in all, you now have a greater possibility to
pinpoint where in the code waits and other latencies occur with the JRA latency capabilities.
z Latency Tabs at a Glance
z Creating a JRA Recording with Latency Data

Using the Latenc y Tabs
z Opening a JRA Recording that Contains Latency Data
z Shared Functionality Amongst All Latency Views
z Using the Latency Timeline Slide Bar
z What is an Operative Set?
z Working with an Operative Set
z About the Event Types View
z Using the Event Types View to Decrease Displayed Events
z Using the Event Types View to Work with Operative Sets
z About the Properties View
z Example of How to Compare two JRA Recordings where one Contains Latencies
z Example Workflow of How to Find Latencies
z Filtering Latency Event Information
Latency Tabs at a Glance

JRA contains three tabs that all show latency data from different perspectives. These tabs are
prefixed Latency and named: Latency Log, Latency Graph, and Latency Traces (Figure 16-1).
Together with these three tabs, there two auxiliary tabs that allow you to turn on and off event
types on the latency tabs and view properties.
Note: Depending on your settings in the Preferences (see Turning on/off Tabs), the latency tabs
may be hidden when your recording does not contain latency information. See Creating
a JRA Recording with Latency Data for information on how to enable latency
information in your recordings.

Cr eat in g a J RA Re cor di ng wit h La te ncy Da t a
Figure 16-1 Latency tabs at a glance
Creating a JRA Recording with Latency Data

You create a JRA Recording with latency data in pretty much the same way you create a regular
JRA recording (see 3. Create JRA Recording). The difference, though, is that you use a different
profile for creating a recording with latency data than when you create the “normal” JRA
recording. If you use the profile with minimal overhead, JRA will not perform a garbage
collection at the end and beginning of the recording, which minimizes the impact on the system
when creating a recording.
For help on the Advanced option for the recording profile, see To use the normal recording profile
with advanced options.
The instructions for how to use the Latency Recording profiles are described in:
z To use the normal latency data profile
z To use the minimal latency recording profile

To use the normal latency data profile

folder with running JVM instances.
4. The Start JRA Recording dialog box appears (Figure 16-2).
Figure 16-2 JRA recording with normal latency profile

Cr eat in g a J RA Re cor di ng wit h La te ncy Da t a
6. From the Select recording file drop-down list, choose Latency Recording Normal.
7. Type a descriptive name for the recording in the Local filename field.
The file is created in the current directory of the JRockit JVM process, unless you specify
a different path. If an old file already exists, it will be overwritten by the new recording.
10. Set a threshold value for Latency threshold. The latency threshold is the duration of the
latency itself. As soon as the latency is longer than that threshold, the data will be saved.
For advance option information, see To use the normal recording profile with advanced
options.
11. Click Start.

the JRA tool.
To use the minimal latency recording profile

folder with running JVM instances.

The JRA Recording dialog box appears (Figure 16-3).

Figure 16-3 JRA recording with minimal latency overhead profile
5. From the Select recording file drop-down list, choose Latency Recording Minimal
Overhead.
Minimal overhead means that the capturing of latency data affects the system in the least
possible way, i.e. it will not disturb or put extra load when recording.
6. Type a descriptive name for the recording in the Local filename field.
The file is created in the current directory of the JRockit JVM process, unless you specify
a different path. If an old file already exists, it will be overwritten by the new recording.

O pe ni ng a JRA R e c o rd i ng t ha t C o nt a i ns L a te ncy Da t a
9. Set a threshold value for Latency threshold. The latency threshold is the duration of the
latency itself. As soon as the latency is longer than that threshold, the data will be saved.
For advance option information, see To use the normal recording profile with advanced
options.
10. Click Start.

the JRA tool.
See also: Filtering Latency Event Information.
Opening a JRA Recording that Contains Latency Data

When a recording with latency data is complete, it is automatically loaded in the JRA tool.
To enable latency data on a latency tab

1. Click on a tab with the prefix Latency.
The Show JRA Latency Perspective window opens (Figure 16-4).
Figure 16-4 The Show JRA Latency Perspective window
2. Select Remember my selection if you do not want this window to open the next time you
click on a latency tab.
3. Click Yes.

Shared Functionality Amongst All Latency Views

The latency tabs have some functionality that they share, such as the Latency Timeline slide bar,
the possibility to select events for the operative set, the Event Types view, and the Properties view
(Figure 16-5).
Figure 16-5 Shared latency tab functionality
These topics explain the shared functionality of the latency tabs:
z What is an Operative Set?
z Working with an Operative Set
z About the Event Types View
z Using the Event Types View to Decrease Displayed Events

Usin g t h e La t enc y Timel ine Sl ide Bar
Using the Latency Timeline Slide Bar

The Latency Timeline slide bar is a universal slide bar for all tabs prefixed Latency. It shows the
entire length of the recorded JRA file. Changing the time span or refocusing the Latency Timeline
slide bar affects all latency tabs in JRA. You can also use the scroll and zoom buttons to refocus
on events within the recording.
The different ways to use the Latency Timeline are described in the following topics:
z To decrease the time span on a latency tab
z To refocus using the timeline slide bar
z To move and zoom using the move and zoom buttons
z To reposition the timeline slide bar
To decrease the time span on a latency tab

z Click and drag the handles on the sides of the Latency Timeline (Figure 16-6).
Figure 16-6 The Latency Timeline decreased
To refocus using the timeline slide bar

z Drag the Latency Timeline into the desired position (Figure 16-7).
Figure 16-7 Refocus on the Latency Graph tab
To move and zoom using the move and zoom buttons

1. Click the move buttons (left or right) to move the Latency Timeline. The scroll buttons are
marked 1 in Figure 16-8.
Figure 16-8 Move and zoom buttons

2. Click the zoom in/out buttons to shorten the time span on the Latency Timeline. The zoom
in/out buttons are marked 2 in Figure 16-8.
3. Double-click the Latency Timeline slide bar to go back to display the full length of the latency
data.
To reposition the timeline slide bar

1. On the Menu bar, select Windows>Preferences...
The Preferences dialog box appears (Figure 16-9).
Figure 16-9 Preferences dialog box
2. In the left panel, select JRockit Mission Control>Runtime Analyzer>Latency.

The Latency panel appears (Figure 16-10).

W ha t i s an O pe ra ti v e S e t ?
Figure 16-10 Preferences dialog box—Latency panel
3. In the Range Selector position box, select the radio button that identifies where you want the
timeline slide bar to appear; for example, if you want move the slide bar to the bottom of the
tab, select Bottom.
4. Select either Apply (if you want to set more preferences) or OK.
Note: For this change to take affect, you will need to close and reopen the particular recording.
What is an Operative Set?

An operative set is a set of events that you choose to work with. You can think of the operative
set as a selection of events that you find particularly interesting to view. If you select events for
the operative set on one latency tab, those events are remembered for the other latency tabs and
you can easily view those events on the specific latency tab by selecting the Show only
Operative Set option (Figure 16-11).
Figure 16-11 Show only Operative Set option

Working with an Operative Set

You can add and delete events in your operative set in different ways depending on which latency
tab you are looking at the moment. The procedures
z To select events for the operative set

and
z To remove events for the operative set

describe how to add and remove events from the operative set. To look at the operative set,
describes how to use the operative set within a tab. The ways on how to work with the operative
set is similar on all latency tabs. These instructions describe how the operative set works for the
Latency Log tab.
For an explanation of an operative set, see What is an Operative Set?.
To select events for the operative set

1. Click on any of the latency tabs, for example the Latency Log tab.
2. Right-click on one or select several events in the Event Table.
3. Select Operative Set > Add selection or Operative Set > Set selection.
The Add selection option adds the events to an already existing operative set (or to a new
one).
The Set selection option clears and overrides the current operative set with the events that
you currently have selected.
To remove events for the operative set

1. Click on any of the latency tabs, for example the Latency Log tab.
2. Right-click on one or select several events in the Event Table.
3. Select Operative Set > Remove selection or Operative Set > Clear.
The Remove selection option deletes the currently selected events from an already existing
operative set.
The Clear option deletes all events from an already existing operative set.

Abo ut the Ev ent T ypes Vi ew
To look at the operative set

z Select the Show only Operative Set on the latency tab that you are at.
Notice that the list of events becomes more manageable.
About the Event Types View

The Event Types view lists the events in relation to where they come from. The Event types
themselves (marked 3 in Figure 16-12) come from a Level (marked 2 in Figure 16-12), and the
Level comes from a Producer (marked 1 in Figure 16-12).
Figure 16-12 Producers, levels, and event types
Below is an explanation of what you can see in the Events Type view (Figure 16-12):
1. Producers are the part of the system that produced the events, for example, garbage collector
and the JRockit JVM. A producer can come from a third party that uses the latency recording
API.
2. Levels are a subdivision of producers. Two events within the same level and thread can never
be performed at the same time. Levels are best visualized when looking at a thread in the
Latency Graph tab (see Understanding the Different Parts of a Thread Image). There you see
that the thread is divided into several levels, but two events within a level overlap.
3. Event type is the actual type of event that was responsible for the latency.

Using the Event Types View to Decrease Displayed Events

If you have many events selected in the Event Types view, the Latency Timeline becomes quite
saturated with information (Figure 16-13). It is a good idea to decrease the amount of events to
eliminate events that are not interesting to view, for example, events that come from the JVM
level.
Figure 16-13 Latency Timeline saturated with information
You decrease (or increase) the amount of data displayed in the Latency Timeline by deselecting
events in the Event Types view (Figure 16-14).
Figure 16-14 The Event Types view
To change the amount of events displayed

1. Click the Event Types view.
2. Click on a specific event, a level, or a producer to select or deselect (see Using the Event
Types View to Decrease Displayed Events for an explanation of producer, level, and event).

Using the Event Type s View to Wo rk with Ope rati ve Sets
The Latency Timeline in Figure 16-14 now looks something like Figure 16-15 when many of the
events have been removed. Notice how much easier it is to see differences over time.
Figure 16-15 The Latency Timeline with events removed
Using the Event Types View to Work with Operative Sets

Read about operative sets at What is an Operative Set?.
You can add events to an operative set or remove it from the set directly from the Event Types
view. This feature is useful when you want to add or remove all events of that type, add or remove
all events from that specific thread, and so on. These features are enabled by using a context menu
accessible from the Event Types view (Figure 16-16).
Figure 16-16 Event Types operative set context menu
The procedures in this topic show you how:
z To select events for the operative set
z To remove events for the operative set
To select events for the operative set

1. Open the Event Types view by opening the Windows menu and selecting Show View >
Operative Set.
2. Right-click a type in the Event Types view (see About the Event Types View for a description
of tab contents).
The context menu appears.
3. Select Operative Set > Add selection or Operative Set > Set selection.
The Add selection option adds the type to an already existing operative set (or to a new
one). Adding the event type to operative set adds all events of the selected type to the set.

The Set selection option clears and overrides the current operative set with the types that
you currently have selected.
To remove events for the operative set

1. Open the Event Types view by opening the Windows menu and selecting Show View >
Operative Set.
2. Right-click an event in the Event Types view.

The context menu appears.
3. Select Operative Set > Remove selection or Operative Set > Clear.
The Remove selection option deletes the currently selected events from an already existing
operative set.
The Clear option deletes all events from an already existing operative set.
About the Properties View

The Properties view lists the event properties, the event’s stack trace, or the general event data
depending on the view you have chosen (Figure 16-17). You select view by clicking on the button
that corresponds to the view you want to see.
Figure 16-17 Properties view
Below is an explanation of what you can see on the Properties view (Figure 16-17):
1. Buttons for choosing the property you want to view.

E x amp l e o f Ho w t o C o m pa re tw o J R A R e cordi ngs where o ne Co nt a ins La ten ci es
Table 16-1 gives an explanation to the different buttons.
Table 16-1 Properties view buttons

Button Description
Event Properties button. Shows the properties of the specific

event. These properties are the same as found on the Event Details
panel on the Latency Log tab.
Event General button. Shows keys and their respective values for
each event.
Stack Trace button. Shows the stack trace for a specific event.
2. List of information. This list changes content depending on the button you click in the tab.
Example of How to Compare two JRA Recordings where

one Contains Latencies
In this example you will see two recordings from the same application. The application that has
been recorded uses a common method for logging transactions, which causes many latencies due
to Java synchronization. These latencies can be found in almost all threads in the recording that
is named pricing-server-logging.xml.zip (Figure 16-18).

Figure 16-18 pricing-serving-logging.xml.zip with latencies
For the second recording the same application has been used, but the calls to the logging system
has been removed, which causes a lot less latencies in the system. The second recording is named
pricing-server-no-logging.xml.zip (Figure 16-19). You see the difference both in the color
scheme and the Latency Timeline slide bar.
Figure 16-19 pricing-server-no-logging.xml.zip with no latencies
You can compare the two JRA recordings next to each other within JRA, which makes it easier
to see what has happened with the changes in the application (see To compare and contrast JRA
recordings for information on how to compare recordings).

Example Workflow of How to Fi nd Latenci es
Example Workflow of How to Find Latencies

The application that has been used in this example contains a common method for logging
transactions, which causes many latencies due to Java synchronization. These latencies can be
found in almost all threads in the recording. This section will guide you through how JRA can be
used to find which method that contains the latency.
Note: This is an example recording that contains extremely visible latencies, the application
that you are looking at might not contain as obvious latencies.
Look in Figure 16-20 for an example workflow of how to start your latency detective work.
Figure 16-20 Example workflow for finding latencies
The workflow is divided into the following instructions:
z 1. Create a JRA Recording with Latency Data
z 2. Open the JRA Recording in the Latency Graph Tab
z 3. Look on the Latency Traces Tab to Find Specific Method
z 4. Add a Suspected Method to the Operative Set
z 5. Look at Operative Set on the Latency Traces Tab
z 6. Perform Changes to Your Application
z 7. Compare and Contrast Recordings

1. Create a JRA Recording with Latency Data

Before you start profiling your Java application, you need to create a JRA Recording with the
latency recording profile (Figure 16-21). See Creating a JRA Recording with Latency Data for
instructions on how to create a recording.
Figure 16-21 JRA recording with normal latency profile
Note: Before the recording finishes, you might be prompted to filter event information from the
recording because the number of events exceeds a defined threshold or your machine’s
capacity to handle all of them. See Filtering Latency Event Information for details.
Investigate further by opening the JRA recording, see 2. Open the JRA Recording in the Latency
Graph Tab.

2. Open the JRA Recording in the Late ncy Graph Tab
2. Open the JRA Recording in the Latency Graph Tab

Open your JRA recording that contains latency data and click on the Latency Graph tab to see an
overview of all threads. This tab offers a great overview of a first glance to find latencies. In
Figure 16-22, the Latency Graph tab is visible and possible latency events from the Java producer
has been selected in the Event Types view. The color of the Java Synchronization event in almost
is visible in almost all threads of the application, which gives a hint that the Java Synchronization
event causes latencies. Investigate further by looking at the Latency Traces tab, see 3. Look on
the Latency Traces Tab to Find Specific Method.
Figure 16-22 Latency Graph tab with Java events selected
3. Look on the Latency Traces Tab to Find Specific

Method
Once you are done viewing your recording from a threads perspective, you click on the Latency
Traces tab to find methods that contain latencies. The Traces table is sorted to show the methods
that contain the most number of events with latencies first. Figure 16-23 shows that the most
latencies are within the method java.util.logging.FileHandler.publish(LogRecord).

Investigate further by adding the method

java.util.logging.FileHandler.publish(LogRecord) to the operative set, see 4. Add a
Suspected Method to the Operative Set.
Figure 16-23 Latency Traces tab with method that contains latencies
4. Add a Suspected Method to the Operative Set

When you have found a method that contains latencies, you can add that to the operative set. By
adding the method to the operative set, you can concentrate your viewing to the pieces of
information that you are mostly interested in viewing even on other latency tabs. Figure 16-24
shows how to add the method java.util.logging.FileHandler.publish(LogRecord) to
the operative set. Notice how the Latency Timeline changes color (the operative set becomes
blue) when you have made a selection to the operative set.

5. Lo ok at Operati ve Set on the Latency T races Tab
Investigate further by looking at the method that you have selected to the operative set in the
Latency Log tab, see 5. Look at Operative Set on the Latency Traces Tab.
Figure 16-24 Adding method to operative set
5. Look at Operative Set on the Latency Traces Tab

The Latency Log tab presents, in a sorted list, events that contain the most latency. In
Figure 16-25 only the operative set is shown and you see that the first event is causing latencies
in Thread-7. Look at the Event Details panel for property and stack trace information.
Now you might have a pretty good idea of where in the code you need to perform changes.
Perform those changes and create a new JRA recording to compare and contrast the results, see
6. Perform Changes to Your Application.

Figure 16-25 Looking at operative set on the Latency Log tab
6. Perform Changes to Your Application

Once you have found which methods and events that cause latency problems you need to perform
changes to your application code. Perform those changes, create a new JRA recording, and
compare and contrast the result, see 7. Compare and Contrast Recordings.
7. Compare and Contrast Recordings

The latency problem has now been fixed in the example application and the result can look
something similar to what you see in Figure 16-26.

Fil tering Latency Event Informatio n
Figure 16-26 Logging method reworked
Filtering Latency Event Information

When making a latency recording with a low latency threshold or making a recording that runs
for long time, several million events can occur. Often, this can exceed what the JRockit Mission
Control Client is able to handle, usually because the computer doesn’t have enough memory or

because it takes too much time to process all the events in the user interface. To prevent problems
from occurring because of too many events being recorded, the JRockit Mission Control Client
allows you to filter out the following types of events:
z Events that were shorter than a certain threshold set in the recording loading wizard.
z Events in a certain time interval; for example, all events from 10 to 30 seconds.
z Events that belongs to certain thread. You can select a set of threads in a list box.
z Events of a certain event type.
z Events randomly given a certain probability.
To filter latency event information

When you are creating your recording and an excessive amount of events occur, the Open JRA
Recording dialog box appears (Figure 16-27).
Figure 16-27 Open JRA Recording Dialog Box
Select the criteria by which you want to filter the events and click Continue.
The report will continue to load, filtering events by the selected criteria.

C H A P T E R 17
Latency Log Information
The Latency Log tab lists the latency events that took place during the recording. By looking at
latency data in the Latency Log tab, you can easily find a specific event type or select an attribute
by using the sort and filter functions.
Note: The latency events that are recorded do not necessarily mean that they cause any
problems in the running of the application.
z Getting Familiar with the Latency Log Tab
z Changing Start Time View on an Event
z About Details for Events
z Selecting an Event
z Understanding Event Details
z Viewing General Event Details
z Viewing Event Property Details
z Viewing Event Stack Traces
z Viewing Event Property Information

L a ten c y L o g I nf o r m at i o n
Getting Familiar with the Latency Log Tab

The Latency Log tab contains Latency Timeline information, an Event Table, and Event Details
(Figure 17-1).
Figure 17-1 The Latency Log tab
The Latency Log tab is divided into the following sections:
1. Latency Timeline slide bar—this timeline shows the entire recording in its full length (the
Latency Timeline works the same on all tabs that start with the name Latency, see Using the
Latency Timeline Slide Bar for more information).
3. Event Table—the Event Table lists all events that took place during the recording.
4. Event Details—this panel lists the most common types on the heap at the end of the
recording.

Ch an gi ng St ar t Ti me Vi ew on an Ev ent
5. Show only Operative Set—this option allows you to concentrate on studying the events that
you have chosen for your operative set (see What is an Operative Set? for a description of the
operative set and how to use it).
Changing Start Time View on an Event

The JRA recording collects data for different start times of a recording, for example, time since
the recording started and since the JRockit JVM started. This section describes how to change the
view of the start time.
To change the start time view

1. Right click the Event Table (or right click on the label Start Time on the General tab under
the Event Details panel).
2. Click Start Time.
Figure 17-2 Start Time selection
3. Select one of the following Start Time views:

– Time since recording started: this is the default view. It shows how much time has
elapsed since the JRA recording started (in seconds and milliseconds).
– Time since recording started in seconds: this is a shorter view where you see how
many seconds have elapsed since the recording started.

– Time since JRockit started: this view is useful when you have created a recording and
used a JRockit JVM that has, for example, been running on your network for a period
of time. This viewing option shows the time divided in hours, minutes, and seconds.
– Time since JRockit started in seconds: this view is useful when you have created a
recording and used a JRockit JVM that has, for example, been running on your network
for a period of time. This viewing option shows the time in seconds.
– Timestamp: this view shows actual time and date for when the event happened (on the
computer that is running your application).
About Details for Events

The Event Table list on the Latency Log tab lists all latency events that have taken place during
the recording provided the latency sampling was enabled during the recording (see 3. Create JRA
Recording for information on how to record latency data).
z Selecting an Event
z Understanding Event Details
Selecting an Event
There are two places you can view the details for an event: on the Latency Log tab panel called
Event Details or in the Properties tab.
z To select an event and view its details under the Event Details
To select an event and view its details under the Event Details
1. Click the event for which you want to view details.
The event specifics are listed in the panel called Event Details (Figure 17-3).

Und er sta nd in g Eve nt Det ai ls
Figure 17-3 Event selected with General Event Details
You can also view the event specifics on the Properties tab
2. Click on the different tabs in the Event Details panel to see different aspects of detail for the
event.
Note: If you select several events, the Event Details tabs show the information for the event
that was selected first.
Understanding Event Details

As described under Selecting an Event, you have two possibilities to view details of an event:
either directly on the Latency Log tab or in the Properties tab. You also have the possibility to
view the details next to each other, for example, view General event details on the Properties
tab and Stack Trace details on the Event Details panel. Either way, the information is the same.
The description in this help depicts how the event details are displayed from the Latency Log
view. The Event Details are divided into the following sections:

z Viewing General Event Details
z Viewing Event Property Details
z Viewing Event Stack Traces
Viewing General Event Details

To view general event details either click the General tab under Event Details or click the General
button on the Properties tab (Figure 17-4). The General events details is an overview of general
specifics for the selected event (Figure 17-4).
Figure 17-4 The General tab for a thread event
To view General event details

1. Select an event in the Event Table.
2. Click the General tab in the Event Details panel (Figure 17-4).
The following information can be found on the General tab. If something is marked N/A,
it means that there was no information for that piece of information during the recording.
– Index—a number that keeps track of each event in the recording.
– Start Time—indicates the time from when the recording was started (default). You can
change the default setting of the start time, see Changing Start Time View on an Event.
– End Time—indicates when the specific event stopped.
– Duration—the length of the event measured in milliseconds.

V i e wi ng Ev e nt P ro pe rt y D e t ai l s
– Thread Name—the name of the thread you are inspecting.

– Producer—the part of the system that produced the thread, for example, garbage
collector and the JRockit JVM.
– Level—Levels are a subdivision of producers. Two events within the same level and
thread can never be performed at the same time. When you click on several event types
within a level, the events appear on top of each other (see Using the Event Types View
to Decrease Displayed Events).
– Event Type—a subdivision of levels. The Event Type corresponds to what you have
selected on the Event Types tab (see Using the Event Types View to Decrease
Displayed Events).
– Event Type Description—a brief description of the event type, for example, Thread
waiting for a JVM internal event.
Viewing Event Property Details

Select an event in the Event Table and view its details on the tab Event Properties tab
(Figure 17-5).
Figure 17-5 The Event Properties tab
Use the filtering function to find, for example, a specific value, see Filtering Information.
See Also
Event Property Details allows you to jump to the method source code. For more information,
please refer to Jumping to Application Source.
Viewing Event Stack Traces

The Stack Trace tab shows all events on the stack that lead up to the event that you are currently
monitoring (Figure 17-6).

Figure 17-6 The Stack Traces tab
To view the event stack trace

1. Select an event in the Event Table.
2. Click the Stack Trace tab in the Event Details panel (Figure 17-6).
See Also
The Stack Trace tab allows you to jump to the method source code. For more information, please
refer to Jumping to Application Source.
Viewing Event Property Information

The Event Property Histogram (Figure 17-7) shows the properties that accompany every event
that occurs in a latency recording; for example:
z For a “File Write” event this is the number of bytes that were written.
z For a “Java Blocked” event, this is the class of the object the thread used as a lock along
with other information.

Figure 17-7 Event Properties histogram
How the Histogram Works

The histogram shows the distribution of these event properties so that you can see the total latency
that occurred when an event property had a certain value. The histogram is useful if you want to
see latencies grouped by lock class. You can then add all the events that locked on a certain class
to the operative set. The graph and trace tab will then show you in which threads the lock
contention occurred or where in the code it happened.

If you are using JRA as an Eclipse plug-in, you can jump from the Event Properties table, the
Stack Trace tree, and the Event Properties Histogram directly to the source code. A feature
called Jump-to-Source allows you not only to see the name of a “problem” class or method
displayed in the GUI, but lets you jump from the displayed method or class name directly to that
class or method’s source, where you can evaluate the code to see what might be causing the
problem. This feature extremely is useful in helping you locate and debug coding errors that are
creating runtime problems for your application.


1. In the Event Properties table, the Stack Trace tree, and the Event Properties Histogram,
right-click the problem method or class to open a context menu.
2. Select Open Method or Open Type (depending upon what you are jumping from).

C H A P T E R 18
Latency Graph Information
The Latency Graph gives you a graphical overview of how the application executes and it is
easy to select events in terms of when they happened and in which thread. You have a possibility
to both zoom in on a shorter time interval and to magnify the threads themselves to better see the
different events that occurred in the thread.
z Getting Familiar with the Latency Graph Tab
z Filtering on Thread Names
z What Does the Threads Chart Contain?
z Magnifying a Thread
z Showing Garbage Collection Backdrop
Getting Familiar with the Latency Graph Tab

The Latency Graph tab (Figure 18-1) displays the Latency Timeline and the Threads graph.

La ten cy Gr ap h Inf or mat io n
Figure 18-1 The Latency Graph tab
The Latency Graph tab is divided into the following sections:
1. Latency Timeline slide bar with Move and Zoom buttons—this timeline shows the entire
recording in its full length (the Latency Timeline works the same on all tabs that start with the
name Latency, see Using the Latency Timeline Slide Bar for more information).
3. Thread list—a graphic representation of all threads in the recorded JRA file.
4. Thread magnifyer slide bar—this slide bar lets you magnify the thread you are studying.
This way you will better see each event within the thread.
5. Show Transitions—Selecting this checkbox causes arrows to appear indicating thread

transitions. See Showing Thread Transitions.

Usin g t h e La t enc y Timel ine Sl ide Bar
6. Show GC backdrop and Show only Operative Set options—the Show GC backdrop
option allows you to see each garbage collection as fine lines behind each thread. The Show
only Operative Set option allows you to concentrate on studying the events that you have
chosen for your operative set (see What is an Operative Set? for a description of the operative
set and how to use it).
Using the Latency Timeline Slide Bar

Depending on how long your JRA recording is, the Threads graph can be quite cluttered to view
in its full lengths due to all events. Therefore, you can refocus and minimize the amount of data
displayed in the charts by using the Latency Timeline. You can also
Use the move and zoom buttons to refocus in the Latency graph
Move the slide bar from the top of the tab to the bottom.
The different ways to use the Latency Timeline are described in the following topics:
z To decrease the time span on a latency tab
z To refocus using the timeline slide bar
z To move and zoom using the move and zoom buttons
z To reposition the timeline slide bar
Understanding the Different Parts of a Thread Image

A thread contains information on the levels and events that have been taking place during the
recording. Figure 18-2 illustrates how a thread looks when it is zoomed in and magnified.
For information on how to zoom in and magnify a thread, see Magnifying a Thread.

Figure 18-2 Magnifying a and zooming a thread
The following information becomes visible when magnifying a thread:
1. The thread itself. This is triggered by the Producer; that is, the part of the system that
produced an event for that thread, for example, the JRockit JVM.
2. The different levels of the thread. These are imaginary levels and depict that an event can only
take place in one level at a time within the same thread.
3. The events that have taken place in the thread. Each event type has its own color (can be
customized). The start and end of each thread event is marked by a thin black line. When you
hover over an event, you will get more information about that event.
Filtering on Thread Names

The Filter thread names field lets you filter our the threads that you are interested in viewing.
The example in Figure 18-3, depicts how it looks when you have typed in Thread-2. The Threads
graph show the threads starting with the number 2 only, which can make viewing easier.
Figure 18-3 Filtering threads

W ha t D o e s th e T hr e a ds C ha rt C o n t ain ?
What Does the Threads Chart Contain?

The Threads chart lists all threads that have been active during the recording. The threads are
quite colorful at a first glance where every color represents an event. Figure 18-4 shows an
example of threads. The garbage collections are located at the top of the list (marked 1 in
Figure 18-4) and each thread is located below its thread group in alphabetical order (marked 2 in
Figure 18-4).
Figure 18-4 Threads list
Each thread in the Threads list contains events. A thread can also contain different levels within
the same thread (see Thread-14 in Figure 18-4 for an example of levels). To see the actual events
with some granularity, you can magnify the thread itself (see To zoom in on a thread by using the
magnifyer slide bar) and decrease the time span of the thread you are monitoring (see To decrease
the time span on a latency tab).
You can also view specific properties for each event as described in About the Properties View
or Hovering Over an Event. You can also get additional information about a specific thread by
hovering over it to open a tooltip.
Correlating Events on Threads

You can easily correlate latency events that occur on non-adjacent threads by using a guide line
that lays over the graph. For example, if you have 10 threads in your application and you want to
correlate the time when a latency event happens in thread 2 (which will appear toward the top of
the screen) with another latency event that happens in thread 9 (toward the bottom of the screen),
you would do the following:
1. On the elapsed time bar at the top of the Thread List, click the point in time for which you
want to correlate latency events.
The guide line will appear (Figure 18-5).

Figure 18-5 Latency events correlation guideline
Guide Line
2. Scroll down to the first event you want to correlate and hover over the intersection of the
thread event and the guide line to display event information (Figure 18-6).
Figure 18-6 Thread information displayed
3. Scroll down to the next event you want to correlate and hover over the intersection of the
thread event and the guide line to display event information
4. To clear the guide line from the graph, simply click the top of it (the black triangle).
Magnifying a Thread
To get a better view at the events within a thread, you will probably need to magnify the thread
you are monitoring. There are two ways to better see events within a thread: magnify the thread
or zoom in on the time span that is used.
This section explains how to magnify a thread:
z To zoom in on a thread by using the magnifyer slide bar
To zoom in on a thread by using the magnifyer slide bar

1. Click and hold the Thread magnifyer slide bar (Figure 18-7).

Showing Thread Transitions
Figure 18-7 Thread magnifyer slide bar

Maximize
Minimize
2. Slide up to magnify and down to minimize the thread size.
Figure 18-8 shows a thread that has been magnified to its maximum size.
Figure 18-8 Magnified thread
3. Slide up or down, using the side scroll bars, to find the thread you want to study.
4. Hover with the mouse over the thread, you will see details for each event in that thread
(Figure 18-9).
Figure 18-9 Magnified event
As you can see, the events appear as large chunks were there are many of the same type
and each event can be hard to see.
Showing Thread Transitions

A latency event in one thread can be associated with another thread. In the JRockit Mission
Control Client this is called a transition and is displayed as a small black arrow. A transition
could, for instance, happen when one thread that calls Object.notify() and wakes up another

thread that is waiting because of an earlier call to Object.wait(). Figure 18-10 shows an example
of this feature.
Figure 18-10 Thread Transition Arrows
z To enable this feature, select Show Transitions in the Latency Graph Tab.
z If you hold your pointer over a transition arrow, a tooltip will appear containing
information about the selected transition (Figure 18-11).
Figure 18-11 Transition Tooltip
Hovering Over an Event

The default tooltip setting for hovering over an event is to display the standard information
(minimized information plus holder thread and lock name). You can change the amount of
information displayed in the tooltip.
To change the tooltip setting

1. Right-click anywhere in the Threads chart.
2. Click Tooltip settings > Verbosity.

Showi ng Garbage Coll ection Backdro p
Figure 18-12 Tooltip setting
3. Select a tooltip granularity.

– Minimum: shows start time, end time, and duration.
– Standard (default): shows start time, end time, duration, holder thread, and lock name.
– Full: shows start time, end time, duration, holder thread, lock name, and stack trace.
Showing Garbage Collection Backdrop

The Show GC backdrop function is a helpful feature that lets you see when and where a garbage
collection occurs. You will get the best visual effect of the garbage collections if you zoom in on
the threads you are monitoring. The garbage collection backdrop lines might otherwise become
more of a light raster in the background than helpful lines.
To turn on/off the GC backdrop lines

z Click the Show GC backdrop option (marked 5 in Figure 18-1) to turn on/off the GC
backdrop lines.


C H A P T E R 19
Latency Traces Information
The Latency Traces tab contains a list of all methods that contain events with latencies. The
method traces with the most latencies are listed first. The Latency Traces table can be
customized to display specific packages, classes, and methods.
z Getting Familiar with the Latency Traces Tab
z Setting Trace Filter
Getting Familiar with the Latency Traces Tab

The Latency Traces tab (Figure 19-1) lists methods with most amount of events and the longest
latencies.

La ten cy T ra ces In fo rma ti on
Figure 19-1 Latency Traces tab
The Latency Traces tab is divided into the following sections:
1. Latency Timeline slide bar—this timeline shows the entire recording in its full length (the
Latency Timeline works the same on all tabs that start with the name Latency, see Using the
Latency Timeline Slide Bar for more information).
2. Trace filters button—this button allows you to add and remove packages, classes, and
methods in the Traces table.
3. Traces table—the Traces table lists the packages and their events. The color coding of the
# Events and Latency columns gives you an overview of which package contains the events
that have the greatest latencies.

Se tting Trace Fil ter
Setting Trace Filter

Using filters is a great way to minimize the amount of data that is shown in the Latency Traces
table. The available trace filter is quite powerful with capabilities to filter on packages, classes,
and methods. You can create your own filter profile.
The Latency Traces tab has a powerful filtering function that allows you to easily filter out
packages, classes, and methods from the Latency Traces table. That way you will get a better
overview of the exact methods you want to study. You can also decide if you want to show or
hide the stack frames that matches the filter.
To add a package, class, or method

1. Click the Trace filters button.
The Trace filters window opens (Figure 19-2).
Figure 19-2 Trace filters window
2. Click either the Add package, Add class, or Add method button.
The Add window opens (Figure 19-3 shows how to add a package).

Figure 19-3 Add package window
3. Type in the prefix of the package, class, or method name, for example com.bea if you are
adding a package, to quickly find what you are looking for.
4. Select the package, class, or method you want to use as a filter.

The selected package now appears in the Trace filters window (Figure 19-4).
Figure 19-4 Trace filter with package, class, and method
5. Select one of the following options:

– Hide filtered stack frames—you will not see the stack frames that matches the
selected filter.
– Show only traces that match filters—you will see only the traces containing stack
frames that matches the selected filter.
– Enable trace filters—turns the filter function on when selected.
6. Click OK.
To remove a package, class, or method

1. Click the Trace filters button.

The Trace filters window opens (Figure 19-5).
Figure 19-5 Trace filters window
2. Select the package, class, or method you want to remove.
3. Click Remove.
4. Click OK.
Note: You can also deselect the Enable trace filters function to disable the filter.

If you are using JRA as an Eclipse plug-in, you can jump from Method Trace tree directly to the
source code. A feature called Jump-to-Source allows you not only to see the name of a “problem”
method displayed in the tree, but lets you jump from the displayed method name directly to that
method’s source, where you can evaluate the code to see what might be causing the problem. This
feature extremely is useful in helping you locate and debug coding errors that are creating runtime
problems for your application.

1. In the Method Trace tree, right-click the problem method to open a context menu.


C H A P T E R 20
Adding Comments and Notes to a

Recording
JRA is equipped with a small text editor where you can add notes and comments about the
recording and your application. These comments will help the JRockit engineering team to
understand what has happened to the Oracle JRockit JVM and your application during the
recording (Figure 20-1).
Figure 20-1 The Notes Tab
To add a note
1. Enter a description of you application in the text field.
2. To save the message as part of the recording, use one of the options described in Table 20-1.

Ad di ng Commen ts an d No te s t o a Rec ord in g
Table 20-1 Saving a recording

To save the Do this...
recording...
Under its original Select File>Save.

name The comments will be saved in a file.
Under a new name 1. Select File>Save as...

The Save as dialog box appears
2. Open the folder into which you want to save the

recording and enter the name you under which you
want to save the recording.
3. Click Save.
4. Close the JRA recording.

Jrockit Mission Control: Oracle Jrockit Runtime Analyzer

Uploaded by

Copyright:

Available Formats

Jrockit Mission Control: Oracle Jrockit Runtime Analyzer

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Jrockit Mission Control: Oracle Jrockit Runtime Analyzer

Uploaded by

Copyright:

Available Formats

BEA JRockit ®

Oracle JRockit Runtime

JRockit Mission Control version 3.0.2 ®

Introduction to the Oracle JRockit Runtime Analyzer

Workflow Description for Creating and Analyzing a JRA

Alternative Ways to Start a JRA Recording

Oracle JRockit Runtime Analyzer iii

General Information in a JRA Recording

Methods and Call Trace Information

iv Oracle JRockit Runtime Analyzer

Garbage Collection Events Information

Java Heap Content Information

Oracle JRockit Runtime Analyzer v

Code Optimization Information

Lock Profiling Information

Start and End Processes Information

Using the Latency Tabs

vi Oracle JRockit Runtime Analyzer

Latency Log Information

Oracle JRockit Runtime Analyzer vii

Latency Graph Information

Latency Traces Information

Adding Comments and Notes to a Recording

viii Oracle JRockit Runtime Analyzer

Introduction to the Oracle JRockit

z How Does the JRockit Runtime Analyzer Work?

z What is a JRA Recording?

z What is the JRA Tool?

z What’s New in the JRA System?

How Does the JRockit Runtime Analyzer Work?

Oracle JRockit Runtime Analyzer 1-1

Figure 1-1 The JRockit Runtime Analyzer

What is a JRA Recording?

What is the JRA Tool?

1-2 Oracle JRockit Runtime Analyzer

What’s New in the JRA System?

Oracle JRockit Runtime Analyzer 1-3

1-4 Oracle JRockit Runtime Analyzer

Workflow Description for Creating and

Oracle JRockit Runtime Analyzer 2-1

1. Start the JRockit Mission Control Client

2. Start Your Java Application

2-2 Oracle JRockit Runtime Analyzer

3. Create JRA Recording

z To use the normal recording profile

z To use the normal recording profile with advanced options

To use the normal recording profile

3. Click the Start JRA recording button.

Oracle JRockit Runtime Analyzer 2-3

Figure 2-2 Start JRA Recording Dialog Box

4. Select the connection you want to record.

2-4 Oracle JRockit Runtime Analyzer

Figure 2-3 Browse for Project Dialog Box

8. If you want, enter a new Filename.

Oracle JRockit Runtime Analyzer 2-5

12. Click Start.

To use the normal recording profile with advanced options

3. Select the connection you want to record.

5. Click Show Advanced Options.