Performance - Slide 1 - Satyajit P PDF

Scalability, Optimization and Tuning in B2B Sterling Integrator
Rashmi Acharya,
Staff Software developer
June 2012
1 IBM Confidential 2011 IBM Corporation

Agenda
Performance Management
Sterling Integrator Quick Overview
Performance statistics management
Managing BP execution
Proper Cache handling and BP optimization
Tuning of Memory Parameters
GC Analysis and Tuning
Troubleshooting DB growth problem
Managing Connection Pools
Tuning aspects on Adapters and Perimeter services
Tuning improvements for EDI, Translation.
SFG tuning improvements.
Analysing and debugging the performance problems. Thread and Heap and GC analysis

Performance Management Overview
What is Performance Engineering?

What are different Performance tuning parameters?
Performance can be related to:

Latency Amount of time required to generate a response to a request (speed)
Throughput Amount of data transferred in a specified time period (volume)
Scalability Ability of the system to grow with increased workloads (additional hardware)
Load Ability of the system to continue performance levels when the amount of work
requested by the system increases.

Performance Management Overview
Tuning aspects were taken into high priority with the increase in demand and
scalability which includes modifying or restructuring the entire application or code
without modifying the basic functionality.
Key requirements prior to performance changes
Understand your performance requirements
Always plan for peaks not average
Performance check
An objective way of analyzing

IBM Sterling B2B Integration Solution
Solution Capabilities
Powers business process orchestrations, high-volume electronic message exchange,

IBM Sterling B2B Integrator complex routing, translation, and flexible interaction with multiple internal systems and
external business partners
A trusted EDI and data translation solution helping you manage your B2B community
IBM Sterling Gentran
Automates business processes and tax compliance, using global integration standards
IBM Sterling e-Invoicing to support buyer and supplier compliance in a single solution. This is accomplished
while
also maintaining conformity with tax regulations for multiple countries
Provides secure connectivity and collaboration with customers and business partners. It
IBM Sterling B2B delivers unprecedented visibility and control over the business processes shared with
Collaboration Network outside companies
Provides comprehensive partner On-boarding, business process management, and

IBM Sterling B2B community support
Integration Services
A variety of mobile solutions that enable your employees to conduct business easily on
IBM Sterling Mobile solutions a wide range of mobile devices while away from the office

Managing Performance Statistics
2009 IBM Corporation

Managing Perf Statistics
Performance Statistics Report helps

you identify and analyze bottlenecks
in your business processes and the
time taken for completion of each
activity.
Enabling this option uses lot of
system resources and can cause the
degradation itself. So do not enable
this feature as a default option.
The report consists of two parts:

1: Business Process Statistics
2:Internal System Statistics

Business Process
Statistics
Name- Name of a business process
Min (ms) and Max(ms)- Minimum or

Maximum time (in milliseconds) it took for
a business process or activity to complete
Invocations: Number of invocations in

the system since the report was
generated.
Pre-Service (ms)-Total time taken by the

business process engine to complete
activities before a BP or service is run.
Post-Service (ms)-Total time taken by

the business process engine to complete
activities after the business process or
service is run.

Name: Name of the system activity, including the following:
Persist a Business Process Step to the Database

Persist a Document to the Database
Acquire a Database Connection
Execute an XPath Statement
Put a Business Process on the Queue
Min and Max time: Min and Max time system activity took to complete. If multiple
invocations when min/max time will be taken here
Average: Average processing time taken for the system activity to complete.
Invocations: Number of times the system activity was invoked since you last generated
the report.
Total (ms): Total time taken since report generation started.

Persist a Document/BP to Database:
The Persist a Document/BP to the Database statistics is helpful in determining database issues and
slow processing issues.
If the Min (ms), Max (ms), and Average (ms) times are increasing. - Database is becoming full, or
that you have a connection leak.
The number of invocations increase - your persistence level is set too high.
The number of invocations is small and the Min (ms) and Max (ms) times are increasing -Persisting
large documents to the database, which can be moved to the disk to save database space.
Assign from an XPath Statement:
Determine how well-structured XPath statements in your business processes.
The Min (ms), Max (ms), and Average (ms) times are increasing. This indicates that the XPath
statement is not written efficiently, and may slow down your process time.
Good practice to write /PurchaseOrder/location/source () instead of //source() , entire process data

will be traversed.

Acquire a Database Connection:
Helps to figureout database issues, resource leaks, or whether you have to increase the number of
database pools.
The number of invocations are increasing. This indicates that your persistence level is set too high,
causing too much data to be stored in the database, or your cache levels are not tuned correctly, causing
the data to be stored and retrieved from the database instead of from the cache.
The number of invocations are low and not increasing. This indicates that either you have a resource
leak that is not releasing previously used database connections to be used by other threads, or you do
not have enough database pools
Put a Business Process on the Queue:

This is helpful in determining if you are using a queue instead of your cache during processing. When
you compare this information with your benchmarks, verify if you notice increasing invocation times. If
yes, it indicates that you are not using your cache efficiently.

Managing BP Execution

Symptoms and causes of poor BP execution
Increase of business process execution time is because of:
The database is full or is receiving too many requests

Improperly designed business processes.
Improperly Tuned Performance Properties.
Disabled few must do service causing Indexing, archiving, or purging activities may not occur.
There are some system cleaning processes should be running to keep the system healthy.
Disabling few of these processes can also cause the problem.
a. ScheduleIndexBusinessProcess
b. ScheduleBackupService
c. SchedulePurgeService
d. ScheduleAssociateBPsToDoc
e. ScheduleRecoveryBusinessProcess
f. ScheduleAutoTerminateService
g. Shedule_BPLinkagePurgeService

Improperly designed BP can be because of :
Improperly tuned cache properties.

Improperly written XPath statements.
Business process persistence levels are set too high
Unnecessary service invocations.
Using older services and adapters instead of newer and more efficient ones.
Unnecessary loops, inside the business process.
Usage of bad query inside the BP to perform database fetch operation
A full database or DB maintainance issues.
Usage of larger lifespan for documents or BPs unless your business demands for it.
Wrong setting for the Document storage type.

How to determine the cause of poor execution
Review the wf.log file inside the install_dir/logs folder for time and date information showing
increase in the execution time
Review the performance statistics report for number of invocations and processing time.
Review the BP activity step which takes more execution time to check for improper BP
design, mode, Xpath etc. Check for any replacement for the activity to perform any better
solution.
Check for any halted, waiting or interrupted BPs.
Check the document storage type for the BPs.
Persistence level set
Lifespan set for the document or BPs.

Persistence level change
Sterling B2B Integrator can persist either all the data or only a small amount of data for each activity
that is processed. The default value for Sterling B2B Integrator is Full persistence. There are three
levels which you can set persistence:
Global property can be changed in /install_dir/properties/noapp.properties_gis_ext.in file. Values are

SYSTEM DEFAULT, PERSISTENCE_WF_NONE ,PERSISTENCE_ERROR_ONLY.
The parameter #persistence_level=<value> can set to
PERSISTENCE_FULL
PERSISTENCE_MINIMAL
PERSISTENCE_NONE
Business Process Change the persistence level during the check-in process. The business
process persistence level overrides the global persistence level.
Activity Level Change the persistence level in the service or adapter. The activity persistence level
overrides the business process persistence level and the global persistence level.

Best Practices
Document storage type

Choosing the right document storage type:
Database: Store documents in the database. Use this option for small documents or when clustering without a
clustered file system.
File System: Store documents in the file system. Use this option for large volume processing, such as documents that
are over 10 MB in size, or when working with a large number of documents within a single business process (including
any sub-processes it invokes).
Persistence level
We recommend to run the business processes (BP) with minimal persistence required to complete. Do not use full
persistence unless your business needs require it, as this consumes more system resources.
Document and BP lifespan

Do not use larger lifespan for documents or business processes unless your business needs require it. If large lifespan is
not the requirement, it is very important to set the lifespan for BP and document accordingly.

Troubleshooting Cache

Types of Cache Memory
Sterling Integrator uses below types of Caches to execute processes.

Soft Reference Cache:
When objects are removed from the memory cache in order to keep the memory cache size
constant, they are moved to a soft reference cache.
It can grow or shrink based on the available memory.
Memory Cache:
Uses the amount of memory necessary to hold the objects in memory at all times.You can
configure this in the noapp.properties file.
Disk Cache:
Objects can be read more quickly from the disk than from the database. When objects are
no longer in the soft reference cache because they have been garbage collected,
The disk Disk cache cache provides a faster access mechanism than object retrieval from
the database.

Based on the memory
allocated to Sterling B2B
Integrator and the number of
CPUs, The Integrator tuning
wizard will allocate disk
cache and in-memory cache
Shadow Cache
When the thread is available for
the BP to run, the object is
retrieved from the memory
instead of the disk.
Distribution Cache
Distribution Cache Maximum
Determines the maximum number
of business processes in queue on
node2. If the number of business
processes on node2 queue
exceeds this number, node1 will not
transfer any more business
processes to node2.

The number of business
processes that ran without
being cached and the number
that are currently in cache.
Cache location is also
specified so that you can
determine the number of
business processes that were
found in the soft reference
cache, in the disk cache, and
in the memory cache.
Statistics which shows

Soft reference Cache
Memory cache
Disk cache

Cache Usage Monitoring
From the Administration menu, select

Operations > System > Troubleshooter. >
cache usage
The Cache Usage report displays the

following information for each cache type:
Cache name Name of the cache
Count Number of objects in the cache
Requests Number of times an object was
requested from the cache, irrespective of
whether it was found or not.
Hits Number of times an object was
requested from the cache, and was found
successfully.
The Cache loader sizes can be changed in
the CacheManager.properties file. The cache
operation uses expiry feature, LRU based
algorithm for the better performance of it.

Symptoms and reasons of inefficient Cache Usage
Symptoms:
Slow-running business processes
Longer completion time for business processes
Increased cache hits reported in the Cache Usage Report.
No Cache hit percentage
Reasons:
Improper tuning of performance properties.
High cache usage for less frequent large objects.
Low cache usage for high frequent small objects.
Resolving the inefficient Cache Usage:

Properly Tune Cache Performance Properties.
Low Cache size Allocation for Less Frequently Used Large Objects
High Cache Allocation for More Frequently Used Small Objects
Compress Cache Contents

Memory Parameters Tuning

Performance Tuning Utility to
edit the performance
configuration using the settings
based on few internal
calculations.
It is stored in the
tuning.properties.
From the Administration menu,
select Operations > System >
Performance > Tuning. After the
change, run the setupfiles.sh
and then start the instance.
Description of the properties
Physical memory to Platform:

Amount of memory allocated for use
in processing operations. Amout of
memory allocated depends on your OS
and JDK selected. If 64bit jdk selected ,
then we can assign to a size of more
than 4GB memory into the SI system.
Number of Server Nodes:

The number of server nodes
installed. NUM_ASI_NODES in the
tuning.properties file.

Number of Container Nodes:
The number of container nodes installed.NUM_CONTAINER_NODE in the property file.
Allocate memory for Active MQ:

Memory allocated to active MQ messaging Service from the physical memory allocated to SI. If it is true .
Default of 512MB will be allocated. If false memory from physical memory will not be allotted.
IS_DEDUCT_MEM_ACTIVEMQ in the property file.
locate memory for SAP Adapter:

Allocate memory for SAP Adapter from the physical memory allocated to SI. If the value is set to true . Default
value allocated is 256MB.
Allocate memory for Command Line Adapter:

Default of 128MB is allocated if we set the property to true.
Physical Memory Allocated to Server (MB)

This value is not same as physical memory allocated to platform. It is calculated as
((((Physical memory allocated to Platform - (Number of server nodes * Minimum memory allocated to the
server nodes)) - (Number of server nodes * Maximum permanent generation size allocated to server nodes))
- (Number of container nodes * Maximum permanent generation size allocated to container nodes))
Memory allocated to the ops server)))
Here 4096- (1*0)-(1*384) (0*0)-128 =3584.

Initial Heap Size (MB):
Controls the initial size of a Java heap allocated to each server JVM.
it can reduce the overhead associated with garbage collection by improving the server response time
and throughput. If a large number of minor garbage collections are happening, try to increase the
value of it. It can be done with Xms parameter
Example: Linux 64-bit: =if (MEMORY_ASI >= 2036, MEMORY_ASI, if (MEMORY_ASI <= 1024, 512,
round(0.5 * MEMORY_ASI)))
Maximum Heap Size (MB):

Controls the maximum size of a Java heap allocated to each server JVM. It can be done with Xmx
parameter.
If this parameter is tuned correctly, it can:
Reduce the overhead associated with the garbage collection and the risk of
encountering an Out-Of-Memory (OOM) condition
Improve the server response time and throughput
Initial Size of New (Nursery) Heap (MB)

Controls the initial size of the new heap area allocated to each server JVM.
It is used with the throughput/parallel collector instead of the JVM built-in tuning the parameter
provides. Default value is 256MB for Linux. It can be done with Xns parameter
Maximum Size of New (Nursery) Heap (MB)

Controls the maximum size of the new heap area allocated to each server JVM.
Linux 64-bit: Linux-64.MAX_AGE=512
Maximum Permanent Generation Size (MB):
Controls the maximum size of the permanent generation heap allocated to the server JVM. It stores
all the class code and class-like data.
The value of the parameter should be large enough to fit all the classes that are concurrently
loaded. Property value is operating system].MAX_PERM_SIZE=384 Mb for Linux system.
Thread Stack Size (KB):

Controls the maximum Java stack size for any thread. This can be determined with Xss parameter.
Each time a method is invoked, a stack frame is created and pushed into the thread stack. At a
minimum, a stack frame contains a methods local variables and arguments. If a thread's actual
stack size reaches beyond this limit, you will get a java.lang.StackOverflowError exception.
Enable VerboseGC:
JVM Verbose garbage collection (GC) statistics are critical. Select True from the drop-down list if
you want to generate verbose garbage collection data for server JVM. GC logs will be present in
noapp.logs for analysis.
JVM Arguments Prefix:

Determines the arguments set as JVM parameters before other JVM flags are set for server JVM. It
can be used to specify the GC ratio, tenured, nursery area sie ratio etc.
Example: turning on the compaction XXfullCompaction or -XXkeepAreaRatio:10 etc.
JVM Arguments Suffix:

Determines the arguments set as JVM parameters after other JVM flags are set for server JVM. For
example, -Xnoclassgc parameter disables dynamic class unloading.

Garbage Collection analysis and Tuning

Layout of Generation heap
Collecting and analyzing GC
statistics help size the different New objects are allocated in the Eden. When the Eden fills
sections of a heap correctly. up, the JVM issues a scavenge GC or minor collection to
move the surviving objects into one of the two survivor or
GCViewer or VisualGC are few tools semi spaces.
where GC statistics can be analyzed
for memory leaks, Objects allocation The scavenge GCs efficiency depends on the amount of
and deallocation stats, Different heap referenced objects it has to move to the survivor space, and
statistics not on the size of the Eden.
-verbose:gc to get the GC output to When the old heap fills up, the JVM issues a Full GC or
noapp.log file. major collection. In a Full GC, the JVM has to first identify
all the referenced objects. When that is done, the JVM
In th HP JVM or in SUN JVM we can sweeps the entire heap to reclaim all free memory
specify:
java -XX:+PrintGCDetails -
XX:+PrintGCTimeStamps
Xloggc:<gcfilename> -verbose:<gc>

GC performance improvement
Few Parameters which may improve the performance w.r.t. GC
-XX:+DisableExplicitGC
-XX:NewSize --controls the minimum young generation size in a heap.
-Xmn<Size> --sets the size of the young generation heap.
-XX:MaxPermSize Stores all classes. This value of the parameter should be large enough
to fit all the classes that are concurrently loaded.
If GC is not able to clear number of live objects from the heap fo a long time, then there
might be a memory leak which can cause OOM in the future.
In a Full GC, the JVM has to first identify all the referenced objects. When that is done, the
JVM sweeps the entire heap to reclaim all free memory. If you see multiple full GC cycles,
the increase the overall heap size.
If you see increased number of small GC cycles , then try to increase the newSize with
Xmn <size>
-XX:NewSize = (0.33333* value of Xms)
-Xmn = (0.33333 * value of -Xmx)

Understanding of memory parameters in the SI system
If the tmp.sh in the /install/bin folder is configured with these memory parameters
install\jdk\bin\java -XX:MaxPermSize=128m -Xss256k -Xms1024m -Xmx3072m --
XX:MaxNewSize=1024m -Xmn512m -XX:SurvivorRatio=4 -XX:+DisableExplicitGC.
-XX:MaxPermSize=128m , which stores all class objects and toString() done on the class
object will be stored here.
-Xss256k specifies the thread stack memory
-Xms1024m is the min heap size
-Xmx3072m is the maximum heap size , it again depends on the physical memory set in the
machine.
-XX:MaxNewSize=1024m is the maximum size of the young generation heap
-Xmn512m is the size of the eden space i.e new generation heap value..
-XX:NewRatio=2 is the default which is talking about the ratio of new/old generation heap
-XX:SurvivorRatio=4 ratio between eden/survivor space ratio.
-XX:+DisableExplicitGC, i.c application invoked system.gc() will not be entertained.

You can change the size of the
database pools to allow more
connections to the database.
Increasing the pool size can solve
the following issues:
Increased volume levels
Data bursts
Connection or cursor leaks.
Default Document Storage:

FS - Stores the document data on
the local file system
DB - (Default) Stores the document
data in the database
defaultDocumentStorageType in
jdbc.properties
JDBC Monitor Page analysis
Types of Connection pool
UI,Archive,Default,
Non-transactional , Local.

Symptoms that can cause a Database growth
Indexing business process is not working properly, causing no data to be flagged for archival or purging.
Persistence levels set to high globally or in BP or on activity causing DB to grow.
Size of the database is not optimized for your processing requirements
Schedule_BPRecovery business process is either not running or failing causing BP to remain in DB in interrupted
state.
AssociateDocsToBP business process is either not running or failing, causing expired documents to not be
associated for being purged from the system.
Resolving the DB issues:

Monitor the Database Usage Report in Sterling B2B Integrator to determine the levels of database usage to
tune them according to the processing need.
Complete the indexing and purging activities regularly.
Optimizing the document lifespan period in the system
Document Persistence has to be set properly such as FS when dealing with large tables like
TRANS_DATA or DATA_TABLE which maintains the BLOB_DATA.
Running the cleanup activities and regular DB maintenance cycle to make sure of rebuild, shrunk
operations on databases.
Rapid growth tables like DOCUMENT, CORRELATION_SET,TRANS_DATA, DATA_TABLE can be moved
to a separate segments or tablespace to improve the performance.

Managing the Connection Pool
Determines the initial number of connections to be reserved in the database connection pool. It
can be set using the initial sie in tuning wiard.
It is a good practice to set the min connection size to 1 instead of setting to large value.
Periodically check the JDBC connection tracker or connection pool report to check whether it is
matching the need. It is recommended to set the maximum value of 256 in the system for all
connection pools.
Keep proper buffer size value for the future demand during high processing load. Once all the
connections are used (maxsize + buffersize), and a new request for a connection comes in, the
new request is denied and an error is written stating that a connection cannot be created and that
the connection pool has been exhausted.

DB Cleanup Processes
ScheduleIndexBusinessProcess
ScheduleBackupService
SchedulePurgeService
ScheduleAssociateBPsToDoc
ScheduleRecoveryBusinessProcess
ScheduleAutoTerminateService
Shedule_BPLinkagePurgeService

ScheduleIndexBusinessProcess
Runs on completed processes

Picks the Bps with ARCHIVE_INFO with ARCHIVE_FLAG = -1
Sets ARCHIVE_DATE to greatest of:
Lifespan set in workflow definition

Any Documents associated with process.
not doc creation.
Sets ARCHIVE_FLAG to:
0 Archive for archiving the documents
2 Purge for purging them
max_business_processes: Default 5000
ScheduleBackupService
Runs on processes with ARCHIVE_FLAG = 0

Makes backup in filesystem.
Sets ARCHIVE_FLAG = 1
Does not change ARCHIVE_DATE
Starts with oldest process set to archive

SchedulePurgeService
Runs on processes that have:
ARCHIVE_FLAG in (1, 2).
ARCHIVE_DATE < now.
Deletes data from tables based on GROUP_ID

Deletes from ARCHIVE_INFO
query.purge_archiveInfoRowLimit.database_name=2000 (default)
Multi-threaded Purge
ScheduleAssociateBPsToDoc
Finds documents that have:
-1 WORKFLOW_ID.
lifespan in the past.
Documents delete when run of AssociateBPsToDocs purge.

Based on Lifespan considerations.

ScheduleRecoveryBusinessProcess
Abnormal process termination due to hardstop or JVM crash makes the database not in
synchronized state.
There are BPs in indefinite state like active, halting or waiting_on_io. Which are not safe or
final states in the system.
The BPRecover attempts to address the problem of how to synchronize the database and
the WorkFlowEngine so as to not impact an newly executing BPs.
BPReportService obtains the list of ACTIVE, HALTING or WAITING_ON_IO from the
database. This set is then compared to the list of threads, messages in the queues
and ActivityData entries (objects in memory that can be associated with an in-flight process).
This is done 3 times with a 10 sec sleep in between each. If a candidate makes it through all
3 checks, then it is considered active hung.

ScheduleAutoTerminateService
The Auto Terminate service is pre-configured and, by default, is scheduled to run each day
at 4:00 A.M.
The service checks for business processes that have been in a specified state for a
specified length of time and then terminates them.
By default, the Auto Terminate service checks for and terminates business processes
that have been in a halted state for over 14 days. You can adjust these settings to suit your
specific business needs.
Overriding the bprecovery.properties Settings in customer.overrides file:
auto_terminate_days=14
num_states=1
auto_terminate_state1=halted
auto_terminate_batch=1000
In BPML <assign to="AUTO_TERM_DAYS" >new_value</assign>

Reorganize, Rebuild, Stats, Shrink on Databases
These are few important steps in the database maintanance cycle

Reorganize
This function enables a user to clean up the database indexes and other details in order to
make the database function more efficiently . It does a re-ordering of index not re-creating
them again, does not collect stats, uses fewer server resources.
Rebuild
Reduce fragmentation issues caused by frequent insert and purge operation on databases,
rebuild of indexes are very important. Creates a re-ordered copy of the index, and replaces
the original with it, collects stats on the index,
Stats
Collects the statistics on the objects of databases. Helps to create a explain plan for each
objects and keeps them in memory.
Shrink
Releases un-used space in the object back to the OS

Few Important SI tables
The following table lists some of the tables defined in Sterling B2B Integrator experience substantial input,
output, and rapid growth.
These tables are candidates for moving to a tablespace or segment that is separate from the rest of the
database.

Tuning on the Adapter Configurations
HTTP Server Adapter:

The HTTP Server adapter enables you to specify the threads range in the http.properties file, It includes a Min Thread
value and Max Thread value.
If the Max Thread value is reached, any additional connection requests fail.
The http.properties file is located in the properties folder under your application installation directory.
Minimum and Maximum number of Jetty threads can be tuned for the better performance of the HTTP server.
FTP Server Adapter:

FTP Server performance can be tuned by changing
Min and Max threads configuration. Default value is 3 and 6. It the number of requests increases then this property values
can be changed.
Transfer Buffer Size which specifies the size of the buffer for transfer operation, this can be changed to improve the
transfer rate. It depends on the perimeter transfer buffer window size set.
Document Storage Indicates the location where body of the request document must be stored. FS will improve the
performance if it is used as an option instead of DB.

SFTP Server Adapter

SFTP Server adapter to enable external SFTP clients to put files into a Mailbox or get files from a Mailbox.
Minimum and Maximum number of threads can be tuned to improve the performance. These threads are used by the
perimeter server to improve the performance.
Transfer Thread Pool Size: Tuning parameter that indicates the number of permanent transfer threads the server begins
with. Once a socket has either been accepted or connected, the socket is registered with a transfer thread. This thread
asynchronously performs all the input and output for the socket. Increasing the value from default of 2 to higher number will
improve the throughput in the system
Channels per Transfer Thread: A tuning parameter that indicates the number of channels available for each transfer
thread.
Document Storage Type : Select where documents will be stored , it can be DB or FS. FS will always high performant
than DB.
Key Exchange value set to false to avoid the exchange of key during multiple get operations.

Perimeter Server
A perimeter server is communication management software installed in a DMZ that manages communication flows between
a perimeter network and Sterling B2B Integrator TCP-based transport adapters. Perimeter servers help reduce network
congestion issues and enhance security and scalability for high-volume environments.
Memory utilization, the amount of data in these buffers can be managed by setting the high and low water mark settings
for both inbound and outbound traffic.
gmm.shedExecutionTime is the time in ms that perimeter services will wait for memory to be released, when the GMM
memory limit specified by gmm.maxAllocation is reached, before closing one or more connections
gmm.maxAllocation Maximum amount of memory to use for all buffering inside perimeter services and associated
adapters. The default value is 384M.
socket.receiveBufferSize This value is used for socket I/O.

Filegateway tuning
Property: fgRouteConcurrentSessionLimit: Number of File Gateway services that can be run concurrently. Maximum
value can be 64.
If you have a high volume of PGP traffic, you can improve your performance by specifying a group for the file gateway.
pgpCmdline2svcname=PGPCmdlineService. The name Command Line Adapter to be used for PGP packing and
unpacking
If you are processing very large files, increase the probe values to avoid timeout conditions.
bpCompletionProbes.1=120
bpCompletionSleepMsec.1=100
If you have very large files that will be processed by PGP, then increase the fgRoutePGPCmdLineSocketTimeout value
from the default value of 4min.
If you have high volumes of FTP traffic, then specify them as a group , like ftpClientAdapterName=FTPClientAdapter
Suppress Duplicate Messages by mailbox.disallowDuplicateMessages=true
Increase the number of steps a business process must complete prior to returning to the queue. Property:
noapp.AE_ExecuteCycle.# where # is the queue number. Higher values will accelerate individual business process where
lower number provides smooth multitasking capability. Batch processing requires a higher number where as interactive
favors a lower number
Increase the number of concurrent threads. Property: noapp.MaxThreads.
Set storage type. As filesystem . Property: filegateway_ui.storagetype=file

Performance tuning on EDI, Translation
Enveloping Service can be improved by changing these parameters in the system:

In the enveloping.property file:
unpersistedStorageType=DB
TRACKING_LEVEL=none
useLocalInMemoryControlNumbers=true
In the translator.properties.in
If you are translating very large files (5GB or larger) or files that contain many levels of looping, you can
get better translation performance by tuning the following
o parameter: maxKeyFileSize.
o OptimizeXPathHelper property is used for optimize common XPath operations by explicitly
retrieving the data from the DOM.
o Use useTracking=false in the jdbc.properties file
o Check whether the Cache size are set to optimal level .
o Check your Heap and memory settings are supported to

Analyzing the performance problems through different tools
There are several tools avalible for analysing the performance problems.
JVM Thread monitoring tool to monitor
Thread contention issues
Slowness issues
Application hangs issues
dead locks etc
State of the system etc.
Heap Analysis tool: to analyze the
OutOfMemoryError Exceptions
Memory Leaks.
The Application Hangs issues
GC analysis tool to analyze
GC statistics for proper heap settings
Application pause time duing garbage collection
Memory leaks issue
AWR report for
- Database performance issue analysis in Oracle DB
- Low performant sqls in oracle db
- DB server performance improvement
Disclaimer
The information contained in this presentation is provided for informational purposes only.
While efforts were made to verify the completeness and accuracy of the information contained in this presentation, it is provided As Is,
without warranty of any kind, express or implied.
In addition, this information is based on IBMs current product plans and strategy, which are subject to change by IBM without notice.
IBM shall not be responsible for any damages arising out of the use of, or otherwise related to, this presentation or any other
documentation.
Nothing contained in this presentation is intended to, or shall have the effect of:
Creating any warranty or representation from IBM (or its affiliates or its or their supplies and/or licensors); or
Altering the terms and conditions of the applicable license agreement governing the use of IBM software.

Q &A

Performance - Slide 1 - Satyajit P PDF

Uploaded by

Copyright:

Available Formats

Performance - Slide 1 - Satyajit P PDF

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

Performance - Slide 1 - Satyajit P PDF

Uploaded by

Copyright:

Available Formats

Scalability, Optimization and Tuning in B2B Sterling Integrator

1 IBM Confidential 2011 IBM Corporation

2 IBM Confidential 2011 IBM Corporation

What is Performance Engineering?

Performance can be related to:

3 IBM Confidential 2011 IBM Corporation

Key requirements prior to performance changes

Understand your performance requirements

Always plan for peaks not average

An objective way of analyzing

4 IBM Confidential 2011 IBM Corporation

Powers business process orchestrations, high-volume electronic message exchange,

Provides comprehensive partner On-boarding, business process management, and

5 IBM Confidential 2011 IBM Corporation

2009 IBM Corporation

Performance Statistics Report helps

The report consists of two parts:

7 IBM Confidential 2011 IBM Corporation

Min (ms) and Max(ms)- Minimum or

Invocations: Number of invocations in

Pre-Service (ms)-Total time taken by the

Post-Service (ms)-Total time taken by

8 IBM Confidential 2011 IBM Corporation

Persist a Business Process Step to the Database

9 IBM Confidential 2011 IBM Corporation

Assign from an XPath Statement:

Determine how well-structured XPath statements in your business processes.

Good practice to write /PurchaseOrder/location/source () instead of //source() , entire process data

10 IBM Confidential 2011 IBM Corporation

Put a Business Process on the Queue:

11 IBM Confidential 2011 IBM Corporation

2009 IBM Corporation

The database is full or is receiving too many requests

13 IBM Confidential 2011 IBM Corporation

Improperly tuned cache properties.

14 IBM Confidential 2011 IBM Corporation

15 IBM Confidential 2011 IBM Corporation

Global property can be changed in /install_dir/properties/noapp.properties_gis_ext.in file. Values are

16 IBM Confidential 2011 IBM Corporation

Document storage type

Document and BP lifespan

117 IBM Confidential 2011 IBM Corporation

2009 IBM Corporation

Sterling Integrator uses below types of Caches to execute processes.

19 IBM Confidential 2011 IBM Corporation

20 IBM Confidential 2011 IBM Corporation

Statistics which shows

21 IBM Confidential 2011 IBM Corporation

From the Administration menu, select

The Cache Usage report displays the

22 IBM Confidential 2011 IBM Corporation

Resolving the inefficient Cache Usage:

23 IBM Confidential 2011 IBM Corporation

2009 IBM Corporation

Physical memory to Platform:

Number of Server Nodes:

25 IBM Confidential 2011 IBM Corporation

Allocate memory for Active MQ:

locate memory for SAP Adapter: