Presentation

Report
IT Analytics and Big Data
Making Your Life Easier
Paul Smith (Smitty)
Service Management Architect
October 4, 2013
© 2013 IBM Corporation
Agenda
 Data Overload
 Search, Predict, Optimize
 How can IT Analytics help?
2
© 2013 IBM Corporation
Every business has 5-10 critical business process and applications. Slowdown or
outage have a direct impact on their profits, revenue, customers and brand equity
Software problem led to two days of
downtime at the largest bank in Europe
has tarnished their image as the most
reliable banking website.
A leading freight company lost $120
million in revenue because IT was
unaware that critical warning
messages were associated with their
key freight delivery application. They
were unable to deliver packages for an
entire day due to downtime.
The Bottom Line: In Today’s World, the App can never go DOWN!!!
3
© 2013 IBM Corporation
Relevant Operations Data is Huge
A Typical Enterprise of 5000 servers with 125 applications across 2 or 3 data
centers generates in excess of 1.4 TB of data per day
Daily Metric Output:
• 250 Mb of event data from 125,000 Events
• 125Mb of endpoint mgmt data from 5K servers
• 12 Gb of performance data for 5000 servers
• 1 Gb of performance for 5000 Virtual Machine
• 8 Gb or Application middleware data
Assumptions: 40% of servers running monitored
middleware
Average 60 metrics each, collected every 15 min
Average PMDB insert 1000 bytes, 40
inserts/server
• 500 Mb Application transaction tracking data for
125 Applications
• 1 Tb Log file data per day
200 Mb average per server (some will be smaller,
some larger)
Example: WAS instances typically produce
400MB-750MB logs/day
• .35Tb Security data collected per day
• 9 Gb Storage Data per day: 175K fiber ports
175 fiber ports,10 metrics per port, collected every 5
minutes, .5KB per port
25K volumes, 10 metrics per volume, .5KB per
volume
5KB*(65K ports and volumes)*12*24 = 9.3 GB/day
• 2Gb Network performance data for Data Center
networks
180x64 port Switches and 4 Routers to manage
physical network.
Data flow of approximately 1TB unstructured data, and .4TB metric data per day,
Scaled to 20K servers, approx 4TB unstructured, 1.6TB metric data
4
© 2013 IBM Corporation
Managing this much data requires Innovation
Not enough Data leads to Disaster
Too Much Data Overwhelms IT
 Too Little: Limit Data Acquisition and risk missing important data
 Too Much: Flood IT Operations and risk missing important data
 Just Right? Today, we use Tools, Best Practices, Process, and Experience to
get just the right amount of data
 Just Right: Analytics Solutions to examine all data, learn what is
important, and escalate critical problems to Operations staff in a timely way.
 Just Right: Analytics Solutions to get to the heart of the problem.
 Just Right: Analytics Solutions to provide actionable insights.
5
© 2013 IBM Corporation
Enabling business transformation through IT Analytics
Predict
with ease
Capabilities
Specialized
Capabilities
Plug & Play
Architecture
Optimize
Search
all your data
to do more with less
Predict
Search
Optimize
Enable predictive and
preventative
operations and
application
management with next
generation behavioral
learning analytics
Accelerate problem
resolution through
rapid analysis of
structured and
unstructured data..
Diagnose application
and infrastructure
issues with expert
advice.
Optimize resource
deployments with
what-if and best fit
planning tools.
Track capacity and
performance of
applications
Visibility
Control
Automation
Integrated suite of capabilities leveraging existing Application
Performance Management, Event Management and Monitoring Solutions
Operational
Environment
Applications
Systems
Workloads Wireless Network
Voice Security Mainframe Storage
Assets
It’s not just performance optimization. We also have to optimize with
license cost and sub-capacity pricing in mind.
6
© 2013 IBM Corporation
Enabling business transformation through IT Analytics
Predictive Outage
Avoidance
Ensure availability of applications
and services
Faster Problem
Resolution
Find & correct problems faster with
tools that determine actions
required to resolve issues
Optimized
Performance
Improved Insight
Enhance visibility into systems
resource relationships while
increasing customer satisfaction
Track, Optimize, and Predict
capacity and performance needs
over time
Perform
Predict
Resolve
Know
• Use learning tools to
augment custom best
practices
• Identify problems quicker with
insight to large unstructured
repositories
• Determine what resources
are interdependent to assess
impact of failures
• Track capacity and performance
of applications and services in
classic and cloud environments
• Leverage statistical
methods to maximize
predictive warning
• Isolate problems quicker by
bringing relevant unstructured
data into problem investigations
• Gain insight into what is
important to your customer
• Optimize resource deployment
with what-if and best fit
planning tools
• Use past maintenance to
predict part failures
• Repair problems quicker with
the right details quickly to hand.
• Decrease customer churn
and acquisition costs while
increasing customer
retention and satisfaction
• Increase utilization of existing
assets
Lower IT Administration Costs with Automated Analytics
•
•
•
•
7
Escalate performance and capacity issues automatically, reducing manual analysis efforts
Reduce manual customization using learning tools that automatically adjust to new normal
Detect and present problems with a proposed resolution, to be able to do more with less
Advice on Risk based automation to automate low risk tasks and escalate high risk fixes.
© 2013 IBM Corporation
IT Operational Analytics
Performance Data
Predictive
Insights
Unstructured Data
Log
Analysis
Avoid Outages and service degradation
through early detection of abnormalities
Identify problems quicker with insight
to large unstructured repositories
Improve insight though the analytical
discover of metric relationships and trends
Isolate problems quicker by bringing
relevant unstructured data into problem
investigations
Reduce root cause analysis by reducing
time to isolate faulty components in
complex infrastructure
Repair problems quicker with the right
details quickly to hand.
“by 2016, 20% of global 2000 enterprises will have IT
operations analytics architectures in place...”- Gartner
8
© 2013 IBM Corporation
Predictive Insights - The Problem
Why aren’t operations teams preventative today?
- Too much data to analyze manually
- Existing analytic techniques, such as standard thresholds, are not up to the task
- They cannot detect problems while they are emerging (before business impact)
- Set threshold too high, insufficient warning before total failure.
- Set threshold too low, too much noise, everything is ignored
If no there is no ‘early detection’ before the outage, operations teams can only
react while outage is already in effect and already losing money...
9
© 2013 IBM Corporation
Multivariate Analytics
Statistical models can discover mathematical relationships between metrics
Internet Banking
Internet Banking
A
Application
ESB
Java / WAS
AIX
RHEL
Oracle
Core Banking
Application
Windows
z/OS
B
C
D
E
F
G
H
I
The extent this can be achieved depends on a number of factors, such as: range and type of data, availability of data,
and stability of environment. Analytics falls back to a single metric if metrics are unrelated.
10
© 2013 IBM Corporation
Example Scenario: Internet Banking Application
Granger based analytics learns the mathematical relationship between metrics
Web Response Time
Internet Banking
A
B
Anomaly Event
Business Impacted
User Requests
WRT Bad
Web Response
Time
C Time
Web Response
Typical Static
Threshold
WRT Good
D
E
User Requests
F
G
Time
Early Warning
• Learns ‘Web Response Time’ has a normal causal
relationship with ‘User Requests’ - WRT gets slower as
user load gets higher.
Leak
H
I
• If this healthy historical relationship breaks down, say due
to a memory leak, an anomaly is raised immediately
• The problem is detected even while WRT service is “good”
Emerging problems can be detected even while service level are good in absolute term
11
© 2013 IBM Corporation
Value Of The Watson Granger-based Analytic Approach
 Learn normal operational behaviour across
the infrastructure, including how metrics
behave together.
 Maximize Advance Warning: Identifies
metric relationship changes that signal a
problem long before traditional thresholds
 Identify problems before you know to look
for them
 Detect service impacts that are not
identifiable by fixed thresholds alone.
 Assists with root cause analysis by
indicating the most offending metrics.
Provides a more intelligent real-time
assessment of data, able to detect
problems as they are emerging
 Reduces expensive and time consuming
false alerts.
12
© 2013 IBM Corporation
Log Analysis – The Problem
Find the right needle in the haystack – QUICKLY!
It’s SLOW!!
404 ERROR
Where do I
start??
Logs,
Traces,..
Core files
13
[10/9/12 5:51:38:295
GMT+05:30] 0000006a
servlet E
com.ibm.ws.webcontainer.ser
vlet.ServletWrapper service
SRVE0068E:
010001100011100001110
011000111110000110001
111111000110011100011
Centralized,
Distributed, Cloud,
Resilient
Architectures
Increase Data Volume
Everything is
“green”
Events
Transactions
Metrics
Config
© 2013 IBM Corporation
1
4
Log Analysis – Key Capabilities
Accelerate problem isolation, identification and repair
 Advanced search and text analytics across large
volumes of data
 Index, search and analyze application, middleware,
and infrastructure data
 Quickly search and visualize application errors
across thousands of log records
Log
Analysis
 Cross index search across logs and documentation
 Integrate log search with existing service
management tooling to gain multiple perspectives
on a specific instance of a problem
14
© 2013 IBM Corporation
Analytics in IT - Capacity Management
Definition from ITIL V3
– ITIL Capacity Management aims to ensure that the capacity of IT
services and the IT infrastructure is able to deliver the agreed
service level targets in a cost effective and timely manner.
– Capacity Management considers all resources required to deliver
the IT service, and plans for short, medium and long term
business requirements.
Sub Processes
– Component Capacity Management
– Service Capacity Management
– Business Capacity Management
– Capacity Management Reporting
15
© 2013 IBM Corporation
Why Capacity Management is important
Helps consolidate and reduce costs
– Reduces HW and labor costs
– Reduces number of physical servers required to run workloads
– Reduce number of required licenses
Helps ensure application availability
– Are any resources overloaded? When will physical resources reach their limits?
– Have there been any significant changes in my environment between two weeks?
– Ensure supply can meet demand
– Ensure business policies are met
Helps optimize resource utilization
– Right size virtual machines
– Identify trends for workload balancing
16
© 2013 IBM Corporation
Use Analytics to Forecast
You already have the data! Use analytics to:
•
•
•
•
•
•
•
17
Forecast resource bottlenecks
Estimate impact of planned business change
Estimate impact of planned outage (ie maintenance)
Discover risky components
Discover hidden limits and potential unstable components
Give input to performance test decisions
Experiment with placement of workloads (cost, license,
performance, etc)
© 2013 IBM Corporation
Thank You!
18
© 2013 IBM Corporation

similar documents