Automation of Event Correlation and

Clustering With Built-In Machine Learning

Algorithms in Splunk IT Service Intelligence
(ITSI)

Ross Lazerowitz | Product Manager, ITSI

Vineetha Bettaiah | Data Scientist, ITSI

09/08/2017 | Washington, DC
 Forward-Looking Statements
During the course of this presentation, we may make forward-looking statements regarding future events or
the expected performance of the company. We caution you that such statements reflect our current
expectations and estimates based on factors currently known to us and that actual events or results could
differ materially. For important factors that may cause actual results to differ from those contained in our
forward-looking statements, please review our filings with the SEC.

The forward-looking statements made in this presentation are being made as of the time and date of its live
presentation. If reviewed after its live presentation, this presentation may not contain current or accurate
information. We do not assume any obligation to update any forward looking statements we may make. In
addition, any information about our roadmap outlines our general product direction and is subject to change
at any time without notice. It is for informational purposes only and shall not be incorporated into any contract
or other commitment. Splunk undertakes no obligation either to develop the features or functionality
described or to include any such feature or functionality in a future release.
Splunk, Splunk>, Listen to Your Data, The Engine for Machine Data, Splunk Cloud, Splunk Light and SPL are trademarks and registered trademarks of Splunk Inc. in
the United States and other countries. All other brand names, product names, or trademarks belong to their respective owners. © 2017 Splunk Inc. All rights reserved.
 You Need an Approach That…
Provides easy and seamless access to all data of any type and volume

1 Delivers
service context
to prioritize investigation

2 Understands
time-based behavior
based on historical patterns

3 Helps you find what’s broken

quickly with human-scale
actionable alerts
 The Three Pillars of Monitoring Data
ITSI needs to be able to handle all of this in order to be “The Backbone of IT
Monitoring”

​Logs Metrics ​Events

 The Three Pillars of Monitoring Data
ITSI needs to be able to handle all of this in order to be “The Backbone of IT
Monitoring”

​Logs Metrics ​Events

 What the Heck is an (IT) Event Anyway?

▶ For the purposes of this talk when we say “Event” we are referring to Events in
the IT sense not the Splunk sense.
▶ Self descriptive message that tells a user that something happened.
▶ Usually contain some sort of title, severity, and description.
▶ Used to determine in the moment health.
▶ Often very noisey.
▶ Think alarm data coming out of tools like Nagios, Solarwinds, APM, Netcool, etc.
 Example Event
Nagios Health Check

1502642822 src_host="splunk_sh-01" omd_site ="SJC"

perfdata="SERVICEPERFDATA" name="check_dhcp" severity="OK" attempt="1"
statetype="HARD" executiontime="0.000" latency="0.000" reason="OK: Received 1
DHCPOFFER(s), max lease time = 600 sec." result="OK"
From NCTA: https://www.ncta.com/positions/internet-of-things
 The Road to ITSI Event Analytics

ITSI 2.1 .conf 2017

ITSI supports Notable ITSI releases Smart
Events. Mode. ITSI can now use
machine learning to
reduce noise in events.

.conf 2016
ITSI releases the Policy
Engine. Users can curate
policies that reduce the
noise in events and take
automated action.
 Splunk ITSI for Event Analytics
Simplify Your Operations With Artificial Intelligence and Service Context

Service Context Artificial Intelligence Scalable Platform

1001001 0101001
0100010 1001101

0110010111000110

1101011101010110
0010011101011000

Find and fix the most Transform IT operations with Get a full view of your IT
important issues machine learning environment

Separate valuable signal Respond collaboratively

Contextualize and prioritize
in noise and simplify operations
Reduce time-to-resolution on Enable IT with intelligence for Share customized insights across the
business-critical services data-driven decisions enterprise to enable business-centric IT
Under the hood
Vineetha Bettaiah
 Objective
To identify all ideal groups of similar events

▶ Input – Set of events

▶ Features – Each event has a number of attribute fields (topological, numerical,
text, etc)
▶ Output – List of ideal groups with characteristic metadata of each group

name = "check_dhcp
Group 1 severity = “OK”
Metadata reason = “OK: Received 1 DHCPOFFER(s)

Group 2
src_host="splunk_sh-01
Metadata statetype="HARD"
 This is not easy
Big O is not your friend

▶ The computational complexity is exponential

• Number of event feature fields = 78
• Number of possible feature values = 4
• Number of possible groupings = 280
▶ An event can belong to many groups

▶ Brute force is not a practical approach for real time data

Approach and
Algorithm
 Step 1- Feature Extraction

src_host="splunk_sh-01" perfdata="SERVICEPERFDATA”
src_host="splunk_sh-01" perfdata="SERVICEPERFDATA”
severity="OK” name="check_dhcp" …...
severity="OK” name="check_dhcp"
host="splunk01" …...
status=“up” severity=“critical” …...

host={splunk01, splunk02, …}status={up, down} severity={low, medium, critical} …...

 Step 2 – Reverse
Pyramid Clustering
 host= host= severity
L1 splunk0 splunk0
status=
… ... =
up
Group of 1 1 2 critical
Feature

host= host= status=up

L2 splunk01, splunk02, … ... ,severity=
Group of 2 Features status=up status=up critical

L3 host= host= status=up

Group of 3 Features splunk01, splunk02, ... ,severity=
status=up status=up critical

...
...
host= host=splunk01
splunk01, ,status=up,
LN status=up, severity=critica
…. l, …
Step 3 – Backtracking
 host= host= severity
splunk0 splunk0
status=
… ... =
up
1 2 critical

host= host= status=up

splunk01, splunk02, … ... ,severity=
status=up status=up critical

host= host= status=up

splunk01, splunk02, ... ,severity=
status=up status=up critical

...

host= host=splunk01
splunk01, ,status=up,
status=up, severity=critica
…. l, …
 host= host= severity
splunk0 splunk0
status=
… ... =
up
1 2 critical

host= host= status=up

splunk01, splunk02, … ... ,severity=
status=up status=up critical
Group with relatively
number of features in
host= host= status=up common
splunk01, splunk02, ... ,severity=
status=up status=up critical

...

host= host=splunk01
splunk01, ,status=up,
status=up, severity=critica
…. l, …
 host= host= severity
splunk0 splunk0
status=
… ... =
up
1 2 critical

host= host= status=up

splunk01, splunk02, … ... ,severity=
status=up status=up critical
Group with relatively
large number of
host= host= status=up features in common
splunk01, splunk02, ... ,severity=
Group with relatively status=up status=up critical
large number of
events
...

host= host=splunk01
splunk01, ,status=up,
status=up, severity=critica
…. l, …
 Let’s run in real time!

RT Search Policy Yes Policy Engine

Applies?

Notable Events No

Smart Mode

host=splun host= host=

status=up,
k01,status splunk02,
severity= splunk01,
status=up
=up critical status=up

Notable Event Groups

Splunk Demo
Presented by Buttercup Splunker
 The principles behind a successful ML feature
This is where the subtitle goes

Explainability Configurability Extensibility

The feature should be

The feature should be The feature should be
easy to configure and
able to clearly explain extensible so that it can
tune by users of
to the analyst any be quickly iterated
different technical
inferences made. upon.
abilities.
 The ITSI way
This is where the subtitle goes

Explainability Configurability Extensibility

Smart Mode is built on
Every group that is
ITSI provides a UI that top of a Machine
discovered includes a
lets non-Data Scientists Learning Engine, on
human readable
tune the algorithm’s which additional
explanation of how it
parameters. “Factors” will be added
came to be.
over time.
 Q&A
Participant name | Role
Participant name | Role