24
Reality for the Capacity Reality for the Capacity Manager Manager Virtualization, Consolidation, Virtualization, Consolidation, Grids, Blades, Utility Grids, Blades, Utility Computing and Other Ghosties Computing and Other Ghosties [email protected] [email protected]

Reality for the Capacity Manager Virtualization, Consolidation, Grids, Blades, Utility Computing and Other Ghosties [email protected]

  • View
    219

  • Download
    2

Embed Size (px)

Citation preview

Reality for the Capacity ManagerReality for the Capacity ManagerReality for the Capacity ManagerReality for the Capacity Manager

Virtualization, Consolidation, Grids, Blades, Virtualization, Consolidation, Grids, Blades, Utility Computing and Other GhostiesUtility Computing and Other Ghosties

[email protected]@metron.co.uk

A Cornish PrayerA Cornish PrayerA Cornish PrayerA Cornish Prayer

From ghoulies and ghostiesFrom ghoulies and ghosties

And long-leggedy beastiesAnd long-leggedy beasties

And things that go bump in the night,And things that go bump in the night,

God Lord, deliver us!God Lord, deliver us!

Vaporware?Vaporware?Marketshare?Marketshare?

GhostiesGhostiesGhostiesGhosties

· Autonomic SystemsAutonomic Systems

· Knowledge Based Information SystemsKnowledge Based Information Systems

· Expert SystemsExpert Systems

· Artificial IntelligenceArtificial Intelligence

Old computers & OSOld computers & OSOld computers & OSOld computers & OS

• Mainframes: IBM & the BUNCHMainframes: IBM & the BUNCH– VM, MVS, 1100, MCP, VME …VM, MVS, 1100, MCP, VME …– CentralizedCentralized

• Minis: DEC and the restMinis: DEC and the rest– Digital PDP/VMS, DG Nova, Wang…Digital PDP/VMS, DG Nova, Wang…– DepartmentalDepartmental

• Micros: Microsoft, Apple and the restMicros: Microsoft, Apple and the rest– DR DOS, MS-DOS, CPM, CCPM…DR DOS, MS-DOS, CPM, CCPM…– DesktopDesktop

Current servers and OSCurrent servers and OSCurrent servers and OSCurrent servers and OS• z Seriesz Series

– z/OS and z/VMz/OS and z/VM

• UNIX open moving back to proprietaryUNIX open moving back to proprietary– HPUX 11v2, Solaris V10, AIX V5 …HPUX 11v2, Solaris V10, AIX V5 …– SCO, SGI, et al …SCO, SGI, et al …

• LinuxLinux– Red Hat, Debian, Suse, Gentoo…Red Hat, Debian, Suse, Gentoo…

• WindowsWindows– 95, NT4, 98, ME, XP, XP64, Longhorn, Blackcomb…95, NT4, 98, ME, XP, XP64, Longhorn, Blackcomb…

• RAID, DAS, NAS & SANRAID, DAS, NAS & SAN• Networks – LANs and WANsNetworks – LANs and WANs

Marketshare?Marketshare?

VirtualizationVirtualizationVirtualizationVirtualization• Virtual OS – VM, MVS and now UNIXVirtual OS – VM, MVS and now UNIX

– Paging and swapping etc to share resourcesPaging and swapping etc to share resources– Partitions (logical) to isolate domainsPartitions (logical) to isolate domains– HPUX on Superdome; Solaris on Sunfire E25k; AIX on e-p5HPUX on Superdome; Solaris on Sunfire E25k; AIX on e-p5

• OS EmulationOS Emulation– eg PC on a Mac, UNIX on a PCeg PC on a Mac, UNIX on a PC– To enable alien domains, Software overhead = performance hitTo enable alien domains, Software overhead = performance hit– Virtual PC – multiple PC on a PCVirtual PC – multiple PC on a PC

• Server VirtualizationServer Virtualization– VMware versions per domain – Windows, Linux, bladesVMware versions per domain – Windows, Linux, blades– Heterogeneous = ultimate goal -Heterogeneous = ultimate goal -

• Data Virtualization – SANData Virtualization – SAN• Network Virtualization – LAN/WANNetwork Virtualization – LAN/WAN

ClustersClustersClustersClusters

• Distributed ProcessingDistributed Processing

• FailoverFailover

• High availability load balancingHigh availability load balancing

MultiProcessorsMultiProcessorsMultiProcessorsMultiProcessors

• AMP: AsymmetricAMP: Asymmetric– Historic, early DEC implementations etc…Historic, early DEC implementations etc…

• SMP: SymmetricSMP: Symmetric– Dominant, typically 4 - 16 - 64 – 128 CPUsDominant, typically 4 - 16 - 64 – 128 CPUs

• MPP: Massively ParallelMPP: Massively Parallel– Vogue… IBM SP2, Cray to Sun E10000 then SunFire Vogue… IBM SP2, Cray to Sun E10000 then SunFire

range, DEC Alpha Wildfire to Compaq to HP …range, DEC Alpha Wildfire to Compaq to HP …

• NUMA: Non-Uniform Memory AccessNUMA: Non-Uniform Memory Access– Vogue with Sequent NUMA Q to IBM (migrate to p), Vogue with Sequent NUMA Q to IBM (migrate to p),

Pyramid “meshine” to Siemens to Fujitsu with Pyramid “meshine” to Siemens to Fujitsu with Amdahl…Amdahl…

Consolidation byConsolidation byConsolidation byConsolidation by

• Server, Application, Data, LocationServer, Application, Data, Location

• Moving workloads to a larger machineMoving workloads to a larger machine

• Moving workloads to logical partitionsMoving workloads to logical partitions

• Moving workloads to virtual machinesMoving workloads to virtual machines

• Moving data to SANMoving data to SAN

• Moving servers to central locationMoving servers to central location

PartitioningPartitioningPartitioningPartitioning

• PhysicalPhysical– Historic; Dynamic if applications quiescentHistoric; Dynamic if applications quiescent

• LogicalLogical– PR/SM and LPARs on MVS/VMPR/SM and LPARs on MVS/VM– Now available on top-end UNIX boxesNow available on top-end UNIX boxes

• Separate development and productionSeparate development and production

• Separate multiple configurationsSeparate multiple configurations

• Dynamic workload balanceDynamic workload balance

RAID , DAS, NAS & SAN RAID , DAS, NAS & SAN RAID , DAS, NAS & SAN RAID , DAS, NAS & SAN

• Redundant Array of Independent DisksRedundant Array of Independent Disks– EverywhereEverywhere

• Direct Attached StorageDirect Attached Storage– For optimum performance on key appsFor optimum performance on key apps

• Network Attached StorageNetwork Attached Storage– For ease of implementationFor ease of implementation

• Storage Attached NetworkStorage Attached Network– For performance, backup and recoveryFor performance, backup and recovery

GridsGridsGridsGrids

• Limited applications historicallyLimited applications historically

• CPU intensive which can be “parallellized”CPU intensive which can be “parallellized”

• Local clusters, campus grids, net taskers, Local clusters, campus grids, net taskers, global gridsglobal grids

• Current examples are:Current examples are:– Search for Extra-terrestrial IntelligenceSearch for Extra-terrestrial Intelligence

– Network for Earthquake Engineering SimulationNetwork for Earthquake Engineering Simulation

– SETI@home [email protected]@home FightAids@home...

• Next? Oracle Grid?Next? Oracle Grid?

BladesBladesBladesBlades

• Rack mounted, bay in chassisRack mounted, bay in chassis• Hot swappable mid or back planeHot swappable mid or back plane• Own processors, memory, storageOwn processors, memory, storage• Own network controller, OS, appsOwn network controller, OS, apps• Shares power, fans, drives, switchesShares power, fans, drives, switches• Aggregate interconnect, cabling etcAggregate interconnect, cabling etc• Simplify infrastructure managementSimplify infrastructure management• $580M in 2003; $1.3Billion in 2004…$580M in 2003; $1.3Billion in 2004…

Utility Computing Utility Computing Utility Computing Utility Computing

• ““On demand” or “adaptive” or “agile” On demand” or “adaptive” or “agile” or “organic” computing needs:or “organic” computing needs:

– Rapid server provisioningRapid server provisioning– Storage and server virtualizationStorage and server virtualization– Network virtualization - route optimizationNetwork virtualization - route optimization– Web services, xSPs, outsourcing, off-shoreWeb services, xSPs, outsourcing, off-shore– Call-in as needed, pay as you goCall-in as needed, pay as you go– Blade technologyBlade technology– Grid computingGrid computing

ITILITILITILITIL

• Not vaporware – real booksNot vaporware – real books

• More “where’s the meat” but thenMore “where’s the meat” but then

• What did you expect in a book?What did you expect in a book?

• Can’t be domain or site specificCan’t be domain or site specific

• Adopted with:Adopted with:– 6sigma, CMMi etc for process6sigma, CMMi etc for process– Cobit/Sox/Basel II etc for auditCobit/Sox/Basel II etc for audit– TQM, EFQM etc for qualityTQM, EFQM etc for quality

ITIL isITIL isITIL isITIL is• The IT Infrastructure Library - booksThe IT Infrastructure Library - books

– Service Support & Service DeliveryService Support & Service Delivery

– Business Perspective, Infrastructure, Development, Service Business Perspective, Infrastructure, Development, Service ManagementManagement

• Good practice for managing IT & standard termsGood practice for managing IT & standard terms• Basis of BS15000, 7799 and ISO 17799 standardsBasis of BS15000, 7799 and ISO 17799 standards• Developed by UK’s OGC (CCTA) in the 90’sDeveloped by UK’s OGC (CCTA) in the 90’s• Metron key contributor to initial DemonstratorMetron key contributor to initial Demonstrator• itSMFitSMF

– The IT Service Management Forum for ITIL usersThe IT Service Management Forum for ITIL users

– Promotes exchange of info & experiencePromotes exchange of info & experience

– GB, NL, B, AUS, ZA, CDN, F, CH/A/D, USA etcGB, NL, B, AUS, ZA, CDN, F, CH/A/D, USA etc

ServiceManagement

OGC & ITIL frameworkOGC & ITIL frameworkOGC & ITIL frameworkOGC & ITIL frameworkHow to Manage

Successful Programs

How to Manage Change

Risk Management

PRINCE2and other OGC

books

Planning to Implement Service Management

Applications Management

The Business

Perspective

ICT InfrastructureManagement

Service Delivery

Service Support

Security Management

People

Processes

ProductsP

ract

ice

BusinessInfrastructure

ServiceApplication

ITSMITSMITSMITSM

• Service SupportService Support– Incidents and Problems (Service Desk)Incidents and Problems (Service Desk)– Changes and ReleasesChanges and Releases– ConfigurationConfiguration

• Service DeliveryService Delivery– Capacity and Service LevelsCapacity and Service Levels– Availability and ContinuityAvailability and Continuity– FinanceFinance

Capacity ManagementCapacity ManagementCapacity ManagementCapacity Management

• Performance ManagementPerformance Management– MonitoringMonitoring– Analysis (& reporting/alerting)Analysis (& reporting/alerting)– TuningTuning– Iterative implementation of the aboveIterative implementation of the above

• Demand ManagementDemand Management• ModelingModeling• Application SizingApplication Sizing• Capacity DataBaseCapacity DataBase• Capacity PlanCapacity Plan

ITIL Capacity Management Inputs and OutputsITIL Capacity Management Inputs and OutputsITIL Capacity Management Inputs and OutputsITIL Capacity Management Inputs and Outputs

InputsInputsInputsInputs OutputsOutputsOutputsOutputs

Sub-ProcessSub-ProcessSub-ProcessSub-Process

Business Capacity Business Capacity ManagementManagement

Business Capacity Business Capacity ManagementManagement

Service Capacity Service Capacity ManagementManagement

Service Capacity Service Capacity ManagementManagement

Resource Capacity Resource Capacity ManagementManagement

Resource Capacity Resource Capacity ManagementManagement

TechnologyTechnology

SLAsSLAs

Business PlansBusiness Plans

OperationsOperations

Budgets…Budgets…

TechnologyTechnology

SLAsSLAs

Business PlansBusiness Plans

OperationsOperations

Budgets…Budgets…

Capacity PlanCapacity Plan

SLA guidelinesSLA guidelines

ThresholdsThresholds

ChargingCharging

Audits…Audits…

Capacity PlanCapacity Plan

SLA guidelinesSLA guidelines

ThresholdsThresholds

ChargingCharging

Audits…Audits…

ITIL Capacity Management LevelsITIL Capacity Management LevelsITIL Capacity Management LevelsITIL Capacity Management Levels

Business CMBusiness CMBusiness CMBusiness CM

Service CMService CMService CMService CM

Resource CMResource CMResource CMResource CM

Capacity Capacity PlanPlan

Capacity Capacity PlanPlan

CapacityCapacity

DatabaseDatabase

CapacityCapacity

DatabaseDatabase

IterativeIterative

Activities:Activities:

MonitoringMonitoring

AnalysisAnalysis

TuningTuning

ImplementImplement

IterativeIterative

Activities:Activities:

MonitoringMonitoring

AnalysisAnalysis

TuningTuning

ImplementImplement

DemandDemand

ManagementManagement

DemandDemand

ManagementManagement

ModelingModelingModelingModeling ApplicationApplication

SizingSizing

ApplicationApplication

SizingSizing

CDBCDBCDBCDB Capacity DataBaseCapacity DataBase

Business volumes - planners, reportsBusiness volumes - planners, reportsWorkload volumes - users, logs, trailsWorkload volumes - users, logs, trails

Platform and middleware statisticsPlatform and middleware statisticsHardware and RDBMS statisticsHardware and RDBMS statistics

Detailed transaction statistics/ARMDetailed transaction statistics/ARMWeb/ intranet/ network trafficWeb/ intranet/ network traffic

ERP/User Application statisticsERP/User Application statisticsSLAs, users, ops, monitors, logsSLAs, users, ops, monitors, logsNew Systems - Developers, usersNew Systems - Developers, users

What is real?What is real?What is real?What is real?

• What you can measure and manageWhat you can measure and manage– Collect, analyse, alarm, alert, advise, publishCollect, analyse, alarm, alert, advise, publish

• OS instanceOS instance– System, user and process metricsSystem, user and process metrics

• WorkloadWorkload– CharacterizationCharacterization– TrendingTrending

• Performance forecastingPerformance forecasting• Capacity managementCapacity management

Reality for the Capacity ManagerReality for the Capacity ManagerReality for the Capacity ManagerReality for the Capacity Manager

Virtualization, Consolidation, Grids, Blades, Virtualization, Consolidation, Grids, Blades, Utility Computing and Other GhostiesUtility Computing and Other Ghosties

[email protected]@metron.co.uk