Automating IT Capacity Management Ingo Schulz Product Manager IT Intelligence SAS International Agenda Introduction The IT Capacity Planning Methodology IT Capacity Planning leveraging SAS IT Intelligence Setting the scene Applying the methodology Summary Drivers Affecting IT Operations Underutilized capital resources Complexity of discrete elements Lack of agility Shifting costs of labor and automation Better business alignment 1
Benefits of Capacity Planning Cost savings and increased effectiveness Deferred expenditure Economic provision of services Planned buying Reduced risk Fact-based decision making Predictable, defendable, and sustainable investments Maximizes IT credibility Value to applications lifecycle Examples Worldwide leading carrier: Cost avoidance after ten months: $5.3 M Projected cost avoidance 2006: $15 - $20 M Leading European telecom company Approval of capacity manager before buying new hard ware Cost avoidance in the first year: $5.1 M The Capacity Planning Environment Best-practices framework Proven methodology Intelligent automation 2
How to Go Beyond ITIL: The Methodology The ITCAP Methodology Services Servers Resources Priority The ITCAP Methodology Tools / data sources Metrics Alerts, thresholds Reports Reporting formats 3
The ITCAP Methodology Usage Reports Representative values from trends Peak periods Characteristic values Seasonality The ITCAP Methodology Forecasting reports The ITCAP Methodology Planning reports Forecasts Business forecast units Business data Service data Technical data 4
The ITCAP Methodology The Capacity Plan Management information Planning reports Financial information Recommendations How to Go Beyond ITIL: Automating the Methodology Setting the Scene: The Environment Company: Online Bank Data center infrastructure: 5 Mainframes, 1000 UNIX and Windows servers, Business service: Online Banking Service Main applications Account management Depot management Savings management Customer care 5
Setting the Scene: The Online Banking Service Account Management WWWA Online Banking Users Internet Web Servers Application Servers Database Server Setting the Scene: The Current Workflow Once a year for budget planning Manually gather the different input data from the most important servers Manually copy the data into Excel Manually transform the data for reporting Manually format the reports Manually estimates the trend Manually evaluate the results Send the reports with annotations to the IT Director Setting the Scope (OBS) Business reason Acquiring new customers for online banking service Service: Marketing campaign on 1.4.2006 (+ 20%) New Online Banking Application on 1.7.2006 (+ 10%) Online Banking Service Priority: Servers: UNIX web and application servers, mainframe CPU capacity on UNIX servers Disk space capacity on the mainframe Critical for market competitiveness 6
The Capacity Database (CDB) Normalization Duplicate checking Summarization Consolidation Assignment Archiving Thresholds IT Management Data Cost information Business volumes Server information Access & Mapping & Validation SAR data Forward schedule of change Financial data Asset data Computer Performance data IT personnel data Business data Configuration data IT Service catalogues Other IT Service support data Device Dependent Metrics Device Independent Metrics 7
What are SPECints? Source: SPEC, Reference database Identifying Thresholds Response Time vs Utilisation - M/M/n 30 Response Time Multiplier 25 20 15 10 5 0 0.00 0.20 0.40 0.60 0.80 1.00 Server Utilisation OBS: Thresholds 4 processors (V440): -80% alert -70% warning 8-processors (V880): -90% alert -80% warning R1 R2 R4 R16 Source: Alan Knight, Techniques and standards Setting the Thresholds SUN Fire V880 8 processors, 7% multi-processor-overhead, SPECint = 625 Total capacity: 625 * 8 * 93% = 4.650 Alert: (Total capacity = 4650) * 90% = 4.185 Warning: (Total capacity = 4650) * 80% = 3.720 SUN Fire V440 4 processors, 5% multi-processor-overhead, SPECint=469 Total capacity: 469 * 4 * 95% = 1.782 Alert: 1.782 * 80% = 1.425 Warning: 1.782 * 70% = 1.247 8
Capacity Management Reporting IT Director as the sponsor and for approval The Capacity Plan Capacity Planner for business capacity management Resource usage reports (medium-, long-term) Resource usage overviews Demand forecasts Time-of-Day-, Time-of-Month-profiles Exhaust alerts Deviation reports Top-N-reports Performance Analyst for resource capacity management Resource usage reports (short-, medium-term) Resource usage overviews Exception reports Resource demand change reports Deviation reports Time-of-Day-, Time-of-Month-profiles Top-N-reports Reporting Formats Reporting standards with uniform Title and header/footer Unique title, filter, generation date Colors and line styles Solid lines for total capacity, actual values Dashed lines for thresholds, Dotted lines for confidence intervals, Labeling Metrics for axis Time-of-Day Profile 9
Computing Characteristic Values OBS: Filter for prime shift OBS: Hour = Compute hourly averages OBS: Day = Compute 75 th percentile of hourly values OBS: Month = Compute Maximum of daily values Hourly Averages of a Specific Day 15 AUG 2005: 75 th percentile of hourly values for 15 AUG 2005 75 th Percentiles of a Specific Month AUG 2005: Maximum of daily values of AUGUST 2005 10
Resource Usage on Monthly Basis Resource Capacity Management Toolbox Top-Activity-Report Resource usage overview Alerts through exceptions 11
The Five-Million Dollar Picture Same server environment after 10 month Five million dollar cost avoidance through consolidation Challenges of Statistical Forecasting Skilled analyst are not available or busy Many forecasts must be generated Frequent forecast updates are required Forecasting model is not known for each time series Calendar events and business drivers (inputs) may influence the forecasting accuracy High Performance Forecasting I Select the time column Select the input duration Any known seasonality? Select the variable for grouping Select the column to forecast 12
High Performance Forecasting II OBS: Result: = Forecasts for all our servers with the corresponding statistical error Integrating Planning Data OBS: Include increased page hits as driver for CPU usage Integrating Business Planning Data OBS: 1 st April 2006: Marketing campaign increases business demand by 20% 13
Integrating Technical Planning Data OBS: 1 st July 2006: New release increases CPU demand by 10% OBS: Insufficient capacity for account management on WWWA Consolidating Servers Objective: Reducing the physical number of servers Use existing resources more effectively Avoid costs for new servers Reduce complexity of the IT infrastructure How to consolidate? 1. Find consolidation opportunity: Which server offers the most headroom? 2. Combine the two workloads by summarizing the workloads 3. Forecast the workloads for a combined environment Identifying Consolidation Opportunities OBS: NWWWJ offers the best consolidation opportunity 14
Combining the Workloads OBS: NWWWJ can hold the additional workload currently NWWWJ WWWA Combined Forecast OBS: NWWWJ can hold the combined workload also in the future NWWWJ WWWA Comparing Scenarios Replacement Scenario Cost Task SUN Fire V880 $50.000 p.a. on 5 years $10.000 Maintenance $10.000 Setup $4.000 Removal old $1.000 Cost 1 st year $25.000 Cost other $20.000 years Cost 5 years $105.000 Recommendation for Capacity Plan: Invest Consolidation Scenario in consolidation of WWWA Task Cost Savings first year: $9.000, five years: ~ $69.000 Maintenance Setup Removal old Cost 1 st year Cost other years Cost 5 years $5.000 $10.000 $1.000 $16.000 $5.000 $36.000 15
The Capacity Plan The Capacity Plan The Dynamic Capacity Plan 16
Summary The IT Capacity Planning Methodology Automating the Methodology Automated consolidation of all input data Automated creation of reports Automated publishing of reports Resource demand forecasting and assessment of other impacts Integration of financial data Communicate through a Capacity Plan 17
Next Steps IT Capacity Management Training Three-hour IT Capacity Management tutorial Three-day IT Capacity Management education course Discovery Workshop Assessment of IT service issues, needs, requirements, priorities Deliverables as summary of findings, review of priorities, business drivers and critical success factors and recommendations for solution architecture and project plan for implementation 53 18