Application of M

1
Application of MS in Capacity Management within
the ITIL Framework

• Sonya Rahmani
• sonya_rahmani_at_sra.com
• Otto von der Hoff
• otto_vonderhoff_at_sra.com

2
Thesis

• Successful evaluation of system capacity and
  performance requirements for any IT system
  requires an integrated modeling approach
• Driving factor behind successful and
  cost-effective capacity management effort
• MS approach should
• Tightly integrate into Information Technology
  Library (ITIL) practices
• Encompass at least two or more predictive
  modeling techniques
• Complement each techniques respective strengths
  and weaknesses to support the validation of
  predicted results
• Tie to systems performance and workload
  monitoring efforts

sonya_rahmani_at_sra.com otto_vonderhoff_at_sra.com
3
ITIL Background

• Best practices framework and set of guidelines
• Challenge is to translate ITIL theory into
  Operational Reality
• Paper highlights
• How best to integrate MS into ITIL
• Tie to following ITIL activities Monitoring,
  Demand Management, Performance Tuning and
  Application Sizing

sonya_rahmani_at_sra.com otto_vonderhoff_at_sra.com
4
Statistical vs. Simulation Techniques

• Statistical techniques
• Trending using Auto-Regressive Integrated Moving
  Average (ARIMA) models
• Analytical model development efforts
• Simulation techniques
• As is model benchmarks current system
• To Be leverage As Is to develop future
  operating conditions

• Statistical Model
• Strengths
• Shorter turnaround time
• Less detailed input data
• Weaknesses
• Higher risk of being less accurate for predicting
  response times and throughput
• Loss of predictive accuracy where future
  behavioral patterns vary substantially relative
  to historical pattern, and
• Inability to deal with queuing and resource
  contention analysis

• Simulation Model
• Strengths
• Capability for more accurate projections of
  system throughput and response times
• Ability to predict/analyze dynamic queuing
  properties and resource contention conditions
• Weaknesses
• Longer turnaround time
• Requires large volumes of detailed output
  performance data

sonya_rahmani_at_sra.com otto_vonderhoff_at_sra.com
5
As Is Model Development

• Combination of statistical and simulation models
  used to build As Is
• Statistical analysis facilitated service time
  characterization in simulation
• Due to lack of performance data, simulation model
  development would not have been possible without
  statistical models
• Also, if statistical used in isolation, could not
  vary response time and correlate to queuing
  behavior over course of day

Production Statistics
Model Simulation Statistics
Queue begins to build in Model
sonya_rahmani_at_sra.com otto_vonderhoff_at_sra.com
6
To Be Model Development

• Leveraged As Is efforts to forecast impact of
  new workloads and develop To Be models
• Analysis addressed
• Expected response times, throughput, and resource
  utilization
• Impacts anticipated to existing workloads

sonya_rahmani_at_sra.com otto_vonderhoff_at_sra.com