220 likes | 376 Views
A TRUE CAPACITY PLAN IS A PROCESS NOT AN EVENT. Joe Bell Senior Staff Regional Systems Engineer Fujitsu Technology Solutions Inc. joe_bell@ftsi.fujitsu.com 816-509-4084. 04/29/03. Agenda. A brief history of Capacity Planning Processes Mainframe days “Open Systems” beginnings
E N D
A TRUE CAPACITY PLAN IS A PROCESSNOT AN EVENT Joe Bell Senior Staff Regional Systems Engineer Fujitsu Technology Solutions Inc. joe_bell@ftsi.fujitsu.com 816-509-4084 04/29/03
Agenda • A brief history of Capacity Planning Processes • Mainframe days • “Open Systems” beginnings • What’s changed? • Why do we need Capacity Planning? • Basic Capacity Plan Process Components • Workload Analysis • Forecasting • Construct base plan and options • Revision and Technology Inputs • Procurement Analysis • Reporting • Refinement
Agenda -continued • Erector Set Theory • Planning • Production • Verification • Implementation • Maintenance • Continuous Process Improvement • A Complete Process -top down • A Shortcut Process -bottom up • Summary • References
A brief history of Capacity Planning Processes • Mainframe days • The Five Stages of Capacity Planning - Pat Artis • Stage 1 - Vendor control • Reporting Level -> Corporate Decision Makers (CIOs, Directors) • Written Documentation/Plans -> Marketing Proposals • Relation to Corporate Budget -> Synchronized to budget cycle for hardware line items • End User Considerations -> Little to none • Continuity -> Long Term Vendor Marketing Plan • Tools -> Rules of Thumb and Vendor Aids/Tools (proprietary) • Stage 2 - Special Studies • Written Documentation/Plans -> Three to Five year Plans • End User Considerations -> Initial Discussions, business overviews • Continuity -> None - One Time Study • Tools -> Rules of Thumb, Vendor Aids/Tools, Ad Hoc (build your own)
A brief history of Capacity Planning Processes -continued • Mainframe days -continued • The Five Stages of Capacity Planning - Pat Artis • Stage 3 - Technician • Reporting Level -> First Level Technical Supervisors • Written Documentation/Plans -> Reams of detailed technical information, graphics, .. • Relation to Corporate Budget -> Typically none • End User Considerations -> Ineffective surveys or none • Continuity -> Ongoing effort • Tools -> Extensive tool development and acquisition • Stage 4 - Organizational Development • Reporting Level -> Technical Services / Operations • Written Documentation/Plans -> One/ two year plans distributed to management levels • Relation to Corporate Budget -> Timed to serve the budget cycle • End User Considerations -> Basic business requirements understanding • Continuity -> SLRs created & maintained, long range strategies developed • Tools -> Continued refinement • Stage 5 - Mature • Reporting Level -> CIO • Relation to Corporate Budget -> Current and future budget inputs • End User Considerations -> Detailed business requirements understanding
A brief history of Capacity Planning Processes -continued • “Open Systems” beginnings • Disjoint acquisition by end users and applications • No single point of budget management • Plethora of platforms for UNIX(es) and even NT • Hardware cheap, just add more to solve CP and PM • Learning new technologies and network techniques • “Life Blood” applications generally still on the mainframe • If any CP, then Stage 1, 2 or 3
A brief history of Capacity Planning Processes -continued • What’s changed? • Movement of disjoint departmental systems to the Glass House • Server Consolidations for platform and management efficiencies • Systems becoming more reliable and dependable • A handful of sophisticated high reliability, high performance Unix platforms (HPUX, AIX, Solaris,..) • Win 2000 and XP handling larger sophisticated applications • LINUX !! • Not safe to just throw resources at capacity shortages • Application, network and storage subsystem complexities • Many examples of failures, just ask SPEs • Just ask some .bombs and surviving .coms
A brief history of Capacity Planning Processes -continued • What’s changed? • Application workloads growing at higher rates on “open” platforms • Movement of “Life Blood” applications to “open platforms” • Mainframes still do an excellent job, but at what cost? • Most vendor R&D directed to “open platforms” and supporting software • Hey, that hardware isn’t so cheap anymore! • Same for the software • LINUX current exception
A brief history of Capacity Planning Processes -continued • Why do we need Capacity Planning? • You can’t control what you can’t measure • You can’t manage what you can’t control • Technology directions • Can Contribute to Planned strategy versus chaos • Budget expenditures • It’s easier to see where you’re going if you know where you’ve been • Instrumentation for critical management decisions • Couple IT resources with Business Requirements • Better support of the corporate business environment • Just in time resource control for reduced expenditures • It’s Addressed in META Group’s “Best Practices for Monitoring Web Applications” • Improve the Total Cost of Ownership (TCO) • Competitive Pricing • Well configured and balanced systems for top Performance • Fewer systems/Processors for simplified operations and Maintenance
Basic Capacity Plan Process Components • Construct a Plan and Work the Plan • Mission and Objectives • Resource types (processor, memory, Disk, IO, Network) • Long Term, Just-in-time • Types of Servers • User Community, Applications • Define Functions and Activities • Reporting • Forecasting granularity • SLAs • Determine Required Resources and Tools • People • Tool analysis and proposed acquisition • budget impact • Construct Project Plan • Define Deliverables • Define Schedules • OBTAIN MANAGEMENT BUY-IN & APPROVAL
Basic Capacity Plan Process Components-continued • Workload Analysis • Define & Analyze current environment resource requirements • Processors, memory, Disk storage, Network bandwidth • System OS, Database(s), subsystems • Major application(s) (80/20 usually works) • Types (OLTP, Background/batch, web enabled) • SLAs • Concurrent users • Business drivers if available • Performance Profile Requirements • Peak versus Average utilization and ratios • response time requirements for servers, storage, network and users • Construct basic models • Statistical correlation and prediction • Analytical, simulation • Can Workload Types be Balanced (CPU, memory , IO , network) • Identify what’s important to the customer/user • Availability • Response Time • Cost • Scalability
Basic Capacity Plan Process Components-continued • Conduct Forecasting Activities • Historical analysis where appropriate • Required for time series regressions • Good to do anyway for information on historical activity • Gather inputs from application development activities • New applications • New application functionality • They probably won’t have a feel for sizing, try similar and box car • Get with Business Units that depend upon IT resources • A well run business activity will have it’s own forecasts • There may be units that will correlate to application resources • Interview Upper Level Management for other information • Utilize statistical or/and Analytical models • Use Statistics for future based on history • Use analytic for future based on current plus new event approximation
Basic Capacity Plan Process Components-continued • Construct a basic Capacity Plan • The Way Things Are • History and current • Annotated Graphs for events • The Way Things Will Be • Forecasts by application/server • Annotated Graphs with events • Requirements on resources due to business and application changes • Options Analysis • By resource you’re covering (server, memory, IO, Disk, Network) • By vendor/architecture within resource • Use those that satisfy all business requirements, stds. • Create an Executive Summary with all important issues • Include a pricing section if you have any cost data • Some organizations will keep this separate • Some will require option pricing in the Capacity Plan
Basic Capacity Plan Process Components-continued • Review Plan with regard to Infrastructure and Technology • System, Subsystem and Network Plans • New component resource deltas • Additional overheads from functional enhancements • Adjust CP options based on Technology or Architecture Plan • New OS type(s), versions • Three tier web enabled apps. • ISCSI, NAS, SAN • Gigabit network growth • Include Disaster Recovery Requirements • Asset Management Review and Option Pricing Adjustments • If no Asset or Procurement Dept for this then get help from Financial Officer • Include summary in Executive Overview
Basic Capacity Plan Process Components-continued • Finish Plan and make reports/presentations • Present to all possible levels of management • Tech management • Application management • business management • Director IT or CIO for final approval • May have to go back and revisit options • May have to perform budget cut activities against the plan • Remember, “work the plan..”, because it’s not made out of concrete • Stay flexible and keep emotion out of the final result. • Capture data for variances against your plan • Use root cause analysis for improvements to next plan • 80/20 or 90/10, because little ROI on final percentage improvements • Time to begin again..continuous cycle
Erector Set Theory • A Manufacturing Approach to CP • Planning • customer requirements, scope and content, schedules, critical success factors, planning aids • Construction • Gather raw materials, create components, test individual components, distribute analysis to concerned parties, combine components • Verification • Component testability, discrete and measurable assumptions, coherency test, rework time • Implementation • Marketing and distribution for customer acceptance • Maintenance • Expect it, planned enhancements, monthly reporting (forecast to actual), periodic revisions • Continuous Process Improvement -zero defect goal- process verification
Erector Set Theory -continued • Underlying Tools and Methodologies • Benchmarking • Customer Interviews • Basic Regression Analysis • Advanced Regression Analysis • Queueing Theory and Analytic Models • Financial Analysis
A Complete Process -top down • Capacity Plan Process • Develop Master Schedule • Construct Base Plan • Collect Historical CPU Usage • Collect Historical Memory Usage • Collect Historical DASD Usage • Build Baseline Models, Best/I • Itemize Application Events • Collect Business Indicators • Itemize Other Events • Size Application Events • Develop Business Units or NFU Forecasts • Size Other Events • Build Consolidated Forecast • Define Current Configurations • Input the Technology Plan • Determine Workload Placement • Recommend Hardware Platforms • Build Growth Models
A Complete Process -top down -continued • Capacity Plan Process -continued • Construct Consolidated Plan • Distribute Base Plan to all CP and Tech personnel • Input Disaster Recovery Plan • Input Financial Analysis • Input Facilities Plan • Input Performance Management Plan • Develop Plan Text • Inputs from Project Implementation Groups • Input Software Support & Maintenance Plans • Input Operations Plan • Input Network Plan • Input Hardware and Software Financial Analysis • Consolidate Financial Impacts • Consolidate Plan Document • Obtain Executive Approval • Distribute the Plan • Conduct process measures and variance analysis
A Shortcut Process -bottom up • Determine what the Plan Needs to Contain and Plan Structure • Ascertain the Inputs for each section and develop the section
Summary • A useable, accurate Capacity Plan Contains many coordinated activities. • A CP is not a one time effort, but a continuous process of refinement and verification and revision. • The depth of a CP will depend upon the amount of assets you are trying to forecast and the accuracy level demanded by management. Construct a basic Capacity Plan Integrate Infrastructure & Tech requirements Forecasting & Modeling Activities Variance Analysis and Refinement Begin Workload Analysis Data Gathering Pricing & Procurement Options Reporting, Presentation & Approvals
References • Carroll, J. R. & Cool, W. S., “Surviving My First Client/Server, Distributed Systems Capacity Study”, CMG94 Proceedings, 1994. • Bell, J.E. & McCord, D.W., “ESTABLISHING A UNIX MID-RANGE CAPACITY AND PERFORMANCE MANAGEMENT ENVIRONMENT”, CMG95 Proceedings, 1995. • Olsen, L., “Erector Set Theory, A Capacity Planning Approach”, KCCMG, 1993. • Howell, J., “Workload Analysis”, Fujitsu Technology Solutions RSE Training, 2003. • Bell, J.E., “The Origination, Evolution and Future of CRM”, CMG89 Proceedings, 1989. • Artis, P., “The Five Stages of Capacity Planning”, CMG85 Proceedings, 1985. • Thompson, G.I., “Six Levels of Sophistication for Capacity Management”, CMG2000 Proceedings, 2000.