Understanding the Performance
Management Process
Understanding the Performance
Management Process
Monitoring Market
Monitoring in Demand 2004 Management tool study Monitoring tools
account for more then
50% of market
Most organizations
have not matured their monitoring environment Missing Process Complete view Linkage of tools/data Priority Category
What priority for monitoring tools of 23 management categories
Succeeding With Monitoring
Must overcome fragmented purchasing behavior Centralization of data Link disparatetechnical silos via shared process Shared problem identification and resolution central to customer satisfaction
2004 Management Tool Deployment Plans
0.0 10.0 20.0 30.0 40.0 50.0 60.0 Singl Tech Silo Sing le B U/Pr oj. Cons olid ated /Op s Outs ourc ers No c lear p atter n Centralization Occurring
Users are demanding monitoring data in context, leading to a desire to centralize information
The Criticality of Processes
0 20 40 60 80 100 Implementation of standard
processes
Implementation of ITIL based processes
Optimizing organizational structure
Purchasing and deployment management technology
Reacting to problems
All Primary
Without Process Most Organizations Fail Process leads to efficiency Efficiency leads to reduced cost Process usage increasing across IT organizations Investment has focused on change and configuration Each process requires an owner
Successful organizations have implemented processes implemented via tools
Monitoring Process
Goal – maximum performance of technology and
quickest response to issues
A distinction between health/availability
monitoring and performance monitoring
h/a = is it alive and limited metrics
Perf. = analysis of data extensive data for
improvement and resolution
Process must be
Reactive – alarms/alerts/events
Proactive - analysis for planning and improvement
Monitoring: A Practical View
Analysis Collect Data Historical Data DefinePolicy Policies GenerateEvent CorrelateEvents Config
Database Correlation Models Capacity Planning Change Mgmt Config Mgmt Incident Mgmt BRM SLM Performance Monitoring
Fault Monitoring Action
Performance Improvement Reporting
Commonly missing linage to incident, resulting in fragmented actions being taken and no tracking
Normalized Monitoring
Analysis Collect Data Historical Data DefinePolicy Policies Config
Database Correlation Models Capacity Planning Change Mgmt Config Mgmt Incident Mgmt BRM SLM Monitoring Action Performance Improvement Reporting
Move event generation and event correlation into
analysis simplifying activities and normalizing fault and performance activities
Monitoring Feeds for Critical Processes
Monitoring Analysis Collect Data Historical Data Policies Correlation Models Change Mgmt SLM Raw data for analysis Processed data storedfor future analysis
Historical data for analysis
Rules for data analysis (base on business policies) Performance and availability data to feed SLM
Rules for data analysis (based on configuration data) Incident Mgmt Processed data generated an incident for escalation Generate an
automated request for change
Analysis is the key to all monitoring
Peering Inside the Analysis Task
Generate Event Correlate Events Collect Data Historical Data Policies Correlation Models Change Mgmt SLM Incident Mgmt Expected Behavior Apply Algorithms Analysis Rule Anomaly? Request Change? Receive Raw Data NO YES YESAlgorithms will determine the depth and type (e.g., failure, statistical, time-series) of analysis
Handoffs
Process boundaries will vary by
company or even within a single business unit or sub IT group
Handoffs to other processes
requires passing of context
The most difficult part of linking
processes is sharing data
Often tools will span multiple
processes
Linking processes leads to increased efficiency and reduced costs
Enter Capacity Management
The planning of capacity needs and validation
of resource utilization as related to the plan Commonly confused with trending (trend
lines)
Modeling is crucial to true planning
Capacity management a cornerstone process
in new adaptive efforts
Highly redundant to have separate capacity
management and performance management
Data should be shared
Capacity management is increasing scope beyond single devices
Process: Capacity Management
Baseline Environment
Characterize Workload Measure and Modify
Workload Model
Produce Workload Forecast
Workload Model
Define Performance and Availability Capabilities Service Level and
Resource Performance Measures Service Level Performance Predictions Cost Analysis Resource Plan Identify Cost Components
Supply Cost Estimates Define Cost Model
Configuration Management Budget Management BRM Performance
Enter Monitoring Tools
A monitoring process is only as good as the
data that is feeding it
Must exert effort to setting or validating
thresholds and actions
Will have multiple monitoring tools
Acceptable as long as integration is understood Most common to have silo’d tools for specialty
hardware
Centralize around common infrastructure or
applications (e.g., all unix, all J2EE)
New efforts have begun to create
performance repository/warehouse for future analysis
The Three Tier Monitoring Model
Data CollectionData Collection
Mid Level ManagerMid Level Manager
Manager of ManagersManager of Managers
Tier 1 Tier 1 Tier 2 Tier 2 Tier 3 Tier 3
• Data gathering for specific
elements or for response time
• Domain specific tools • Some consolidation
• Goal to minimize vendors for
like technology
• Centralization of alerts and data
for a specific silo
• e.g., NOC, application • Correlation and analysis occur
for the specific domain
A fourth layer emerging for reporting
• Enterprise wide centralization of
key alert data
• Enterprise wide correlation,
notification, and escalation
• Can be an event view, business
view or service level reporting
Numerous tools
One per domain
Assessing Monitoring Maturity
Self assessment
Identify gaps in data collection
Have vendors been minimized?
Is data from elements and response time
being integrated
Real time? Historical?
Is it leveraged for other functions?
Is end user perspective accounted for? Is data being correlated?
At how many levels?
How are results being leveraged?
Does an enterprise view exist?
Bottom Line
Success with monitoring
investments will increase with a investment in monitoring
processes
Ensure processes are linked to
other appropriate operational activities
Share data across functions
Understand the tool deployment
logic and avoid overlap and unnecessary investment