More for less adammetron co uk More for

More for less adam@metron. co. uk More for less in a lean, mean, green, virtualized ITSM world using capacity management best practice it. SMF Norway Sep 2010

Summary 1. 2. 3. 4. 5. 6. it. SMF Norway Sep 2010 More systems Less resources Lean, mean & green Virtualised ITSM world – SDLC & ITIL Capacity Management best practice • Objectives; tasks • I/O; CDB; activities • Deliverables • Monitor & analysis; publish, trend & advise • Capacity Plan • Mapping; distributions & queues • Forecasting; modelling example • CMMI; gap analysis • Pyramid of needs.

# 3 of 30 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 Survey – Do you use these acronyms? • ITIL, ITSM, it. SMFI, it. SMF UK, ITSML library IT infrastructure, Service Management, Van Haren • CMG, UKCMG, Euro. CMG, CECMG, CMGAE, CMGI… Computer Measurement Group • SDLC, BPR Software dev life cycle, Business Process Re-engineering • CMMI, Six Sigma, Balanced Score Card, TQM, TQFM Maturity model, lepto-kurtosis, process quality, ISO 9000 • Sox, Cobit, Basel II Process audit and governance • BCS, NCC, ISO/IEC 20000 Good practice, ethics, standards, certificates, pin badges.

# 4 of 30 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 More systems that have more: • Critical business requirements – IT service and web front end essential to delivery • Users - Business customers, end-users and clients • Services - Business customers’ view of applications • Architectures - abstraction, virtualization, consolidation • Tiers & pools – Web GUI, message handler, app server, warehouse… • Applications – Developments in-house, off-shore, packages… • Servers – Machines, nodes, CPUs, RAM, networks, storage • But what is actually achieved in a business sense?

# 5 of 30 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 Less resources and less: • • Overhead – IT infrastructure and services Finance –budgets cut Sites – consolidated data-centres Staff – cutbacks Spare capacity – headroom, duplex, DR, non-stop Physical servers – virtualised and consolidated Expertise – less experience and less training ‘Inefficiency’, but at what cost to: – Consistent, effective provision of essential services – Increased risk of performance crises.

# 6 of 30 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 Lean, mean & green • • • • Lean - Focus on key objectives and core activities Adopt 80/20 rule – identify busy services/users/servers Exploit spare capacity, minimise headroom Identify how much more traffic can be supported safely Define and agree measurable SLA performance criteria Virtualize infrastructure to avoid procurement by project Mean - Optimise on current investment/reduce licences Green - adopt solutions that are environmentally neutral Centralise to minimise on wastage & reuse heat output Use fewer, more powerful servers to reduce aircon needs Reuse old machines – cascade to good causes Grid and cloud computing… Capacity on demand, use power when really needed…

# 7 of 30 Virtualised Overhead Difficulty of creation & frequency (Hz) Software-only: VMware GSX MS Virtual Server OS-virtualization: ESX, Xen, Hyper-V, paravirtualisation, grid, sysplex LPARs (HP, IBM, Sun) Fairshare, timeshare Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 z/VM, hyperthreading (Intel VT, AMD V’n, IBM Power. PC Software Microcode Hardware

# 8 of 30 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 SDLC – Development Life Cycle Systems or service or Software Deliverables Feasibility Requirements TOR, Scope, PID, SOR Analysis Architecture Functional Spec Design SPE, System spec Implementation Program spec, modules Testing system, a, b, load, trials Deployment Maintenance pilot, production, fix, retire Steps: Review, Feedback, Prototype, Cyclic Routes: Waterfall, Iterative, Scrum Priorities: Project versus Process, Panic versus Procedure.

# 9 of 30 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 ITIL V 2 & V 3 Processes & Functions Intro S Strategy S Design Transition Operation Improv’t Financial M Capacity M Change M Problem M 7 step CSI Demand M Availability M Evaluation Incident M S Reporting Continuity M Validation Event M S Measure Security M Planning Request M Supplier M Release M Access M SLM Configuration SPM/SCM Asset M Process? Knowledge M Service Desk In V 2 & V 3 New in V 3 Changed in V 3 IT Ops M Application M Technical M Function?

# 10 of 30 Six top entities/processes (6 steps) Assets Finance Efficiency Service Port/cat Contracts Availability, Capacity, CMDB, CIS… SLA, OLA, Continuity… CDB/CMIS budgets… Do it Change Start, deliver, stop RFC, CAB, Release… Operations service Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 N Bug? Repair Y

# 11 6 steps applied to all of ITSM of 30 Assets Efficiency Do it Finance Change Bug? Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 ITSD Management of: ITSS/O Management of: Performance & Capacity Incidents Service Levels Problems Continuity Events Availability Configuration Supplier Security Finance… Access…

# 12 of 30 Capacity management objectives “The provision of a consistent, acceptable service level at a known and controlled cost” Objectives Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 Right level of service performance Right level of investment & resource Resources matched to business need Resolve bottlenecks, evaluate tuning Forecast workload demands Evaluate upgrade plans Ensure timely procurements Ensure effective SLA Plan for growth, new apps, new sites Provide performance assurance Deliverables Performance and capacity outputs Demand management criteria Business Workload forecasts Service workload components Event alarms & thresholds Resource usage by domain Capacity plans and acquisition plans SLA targets and measures Application sizing assessments Performance test reviews

# 13 Capacity Management tasks of 30 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 • Core – Past performance trending & reviews – Current analysis, alerting & reporting – Future capacity planning & publishing • Related – Diagnostic monitoring, tracing & accounting – Load balancing & system tuning – Domain expertise and application – Program optimisation – Business requirement liaison – SPE, application sizing, load testing – Service Level Agreements and management.

# 14 of 30 Capacity Management I/O Technology SLA, SLR, OLA Business plans IT plans Deployments Development Operations Budgets… Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 Outputs Inputs Business Capacity Management Service Capacity Management Performance reports Capacity Plans Baselines & profiles Bottlenecks, patterns SLA guidelines Thresholds, alarms New app sizing Audit, costs & charges Outcome Resource/Component Capacity Management Control Throughput Feedback Capacity Database Economy Standards CDB/CMIS Goals Governance

# 15 of 30 Capacity management activities Business CM Service CM Resource/ Component CM Performance activities: Monitoring Analysis Tuning Implement Effective use of deliverables = outcome Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 Performance Reports and Capacity Plan(s) to intranet/email/ pager/paper Reports (regular, ad hoc & exception) to and from: SLM Thresholds Alarms Events Charging Audit Demand Management: priorities, quotas, chargeback Modelling workload forecasting and characteriz’n Application sizing, performance testing and performance engineering Liaison with SLM & Development, test & QA, Business: metrics, plans, KPIs Capacity Database Business and workload volumes Platform and middleware statistics Virtualization statistics Hardware and RDBMS statistics Detailed transaction statistics/ARM Web/ intranet/ network traffic ERP/User Application statistics SLAs and new Systems sizing

# 16 of 30 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 Monitor/analyse

# 17 of 30 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 William Playfair’s dual axis histogram 1821 Early use of correlation – wages vs. wheat over 250 years

# 18 of 30 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 Publish, trend, advise

# 19 of 30 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 Florence Nightingale’s polar pie-chart 1857 Crimean “coxcomb” in proportion to radius not arc

# 20 Mapping business needs to resource demand of 30 Business Qo. S, BMI Weight Dashboard Drivers SLA Service KPI Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 Model Plan Trend/Alert Publish Service Applications(s) Transaction(s) App View Server(s) S Metrics Instrumentation CPU RAM I/O N/W etc

# 21 of 30 Distributions & queues • Nature avoids normal distributions • Computers abhor normal distributions - Normal as derived by Gauss/de Moivre • Computers like exponential distributions - Typically a minimum value and a long tail - Service times are typically exponential - Inter-arrival gaps typically exponential • Under typical enterprise conditions: Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 - Large populations & no priority/ class mix - N queue server, FIFO and exponential applies - “M/M/N” analytical models apply - Utilization U and Service Time S - Response Time R = S/(1 -UN) • And, in simple cases, if U=X% you will wait about X% of the time

# 22 Knee ? of 30 Response Time N > 20 or so N=5 N=2 N=1 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 Service Time Utilization 0 0. 5 1. 0

# 23 Forecasting options: mix and match Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 More potential accuracy and cost of 30 Trust the Vendor Benchmarking and application trials Measurement & Analytical Modeling Outsource it Workload generation and trials Discrete event Simulation Trending Behaviour Patterns Rules of thumb Do nothing More Forecasting Effort

# 24 of 30 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 Modelling example

# 25 of 30 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 Charles Minard’s plot of Napoleon’s campaign Tufte’s “top informative graphic” 1812 Russian campaign graphic done in 1869 Early use of line width to indicate bandwidth

# 26 of 30 The Capacity Plan –Sections • • Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 • • Management summary Introduction Assumptions Business scenarios Service summary – List and brief description of business services supported – Current and recent service provision and resource usage – Service forecasts of business volumes – Corresponding list of workloads & resource requirements – Workload Forecast Scenarios Resource summary – Forecasts - account of impact of workload requirements Options for service improvement Capacity plans should be: – Description of options for required upgrades Short and clear Cost model - quantifies finances for upgrades Ad hoc, periodic, annual? Per service, server or enterprise? Recommendations Based on Qo. S, BMI, KPI or OLA? – Business benefits to be expected Maximum 40 pages per service? – Potential impact of carrying out recommendations Max 2 page management summary? – Risks involved Right outputs on time to right people? – Resources required Targets met and metrics measured? Produced as required and actioned? – Costs, both set-up and ongoing.

# 27 of 30 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 CMP per app per site per stage # CMMI ITSM Cap. Man Task % 5 Optimised b. ITa Business level Dashboard, CPM 2% 4 Measured ITSM Service level SLAM, cap plans 10% 3 Proactive Center Resource level J CDB, Trends, web 30% L 2 Reactive Tickets Analysis Utilization, uptime 55% 1 Ad hoc Help calls Monitor 2 Ad hoc alerts 3%

# 28 of 30 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 Summary of gap analysis procedure • Agree scope of study and checklist of targets – Enterprise Infrastructure, Service/App Portfolio… – Mainframe, warehouse, UNIX, Linux, Windows… – LAN, WAN, SAN, PAN, RDBMS, ERP… • Define objectives, constraints, resources available • Assess versus relevant checklists for Best Practices • Review actual Capacity Management Practice (CMP) • Talk to the Capacity Management Team(s) (CMT) • Readily available reports collected for analysis • Interviews within a week to assess current activities • Review available evidence/ deliverables • Reveal gaps, identify SWOT and next steps • Snapshot report in a week for total focus on material.

# 29 of 30 Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 Risks, costs & benefits for CMP • Benefits • Risks – Improved service provision at cost justifiable service quality – Over expectation of IT by customers – Services that meet Business, Customer and User demands – Too much hardware vendor influence – Increased efficiency – cost savings – Lack of business information Deferred expenditure (cash flow) – Lack • of human upgrade resources • Consolidation (maintenance/licences) – Complex environment • Costs • Planned acquisition (discounts) – Reduced risk – Procurement of required tools • Fewer performance disasters (customers) – Project management as required Fewerincluding performance problems (users) – Staff • costs recruitment, training • Longer Application Life-cycle (fewer abandoned apps) – Accommodation etc • Learning from previous experience, Good – Liaison and interface with business, dev, QA etc. Practice

# 30 Pyramid of needs Potential benefit to enterprise Full Control Prediction Optimal usage of available resources Basic pre-emption of Problems Plan App size Model Demand Tuning Analysis Monitoring & Basic Control D Ca ata Ale ptur rtin e g Capacity Management for Virtual Systems it. SMF Norway Master Sep 2010 Class V 001 Tre An ndin Re alys g po rtin is g A Pla dvic Mo nni e de ng llin g of 30 Acquisition of relevant metrics / Context related knowledge Disorder / Lack of control Monitor CDB

More for less adam@metron. co. uk More for less in a lean, mean, green, virtualized ITSM world using capacity management best practice it. SMF Norway Sep 2010