gostessstlegeB@
`
`4
`
`ANDREW HILLIER
`
`CiIRBAINC.
`
`COPYRIGHT © 2004-2006, CIRBA INC. ALL RIGHTS RESERVED
`
`VMware, Inc.
`
`Exhibit 1012 Page 1
`
`VMware, Inc. Exhibit 1012 Page 1
`
`
`
`

`

`CIRBA SERVER CONSOLIDATION WHITE PAPER
`
`CONTENTS
`
`Tah (ole [0 (e116) aneere 3
`IT Asset Optimization Define «02... ceeececcccececceeeeececeeeceeeeeceaeeecaeeecaceceeeeseaeeeseaeesceeeseuseeteieeesneeeeaes3
`Traditional Approachesto Asset OptimiZation............:ceccceceecceceeeceeseeceeeeeceeeeseaeeeeeeeeteeeessneeesneeenees4
`A Quantitative and Statistical Approach to Asset Optimization .........ccccccccecccecsneeeessteeeeessneeeeseaes 6
`Server Consolidation Analysis OVe@rvieW. ..........:.cccccccceccecseeeeeeeeceeeeeceaeeeeeeeeeeeeesaaeeeseseseeeeetseeeseneesenees 6
`How Consolidation Analysis iS USCC ............ccccccceceesceseeeeeeeeeceeececeeeeeeeeeseaeeeseaeeeseeeeeeeeeseieeesneensseees 6
`How Consolidation Analysis WOrkKS.........cc:cccceccecseeeeeeeeceee cece eeceeeeceeeeceaeeeseaeeeceeeteneeseieeetieenseeees 7
`Factors Affecting Rule Sets ..........cecccececeeesececeeececeeeeeeeeeceaeeecaeeesaeeseeeeesaeeeccaeesseeeesereeseeeesseesenees 11
`Factors Affecting Workload ANnalySis ........0c::cccceececettreeeeeeteeeeetiieeeeetineeeeetiieeeeetiieeeeetiieeeeetiieeeereaa 12
`Other Technical Analysis Factors ..........cccccccccceeeccecseeceeececeeeeeceeeeceeeeceeeeseaeeesceeeceessareessueeesieeenees 13
`Non-Technical AnalySiS Factors .........eccccccceeecceeceeeceeeeeeeeeeeseeeeeeeceeaeeeseeaeeeseeaeeeseeaeeeseenneeseenanees 14
`Benefits of a Quantitative and Statistical ADDrOACh ....... cee cceecceceeeceeeeeceeeeeceeeeeseeeseceeteteeteaeesenees 14
`Realization of Cost RECUCTIONS ..........ecccecesceeeececenececeeeeeeeeeeeececeaececaaeceeeeecaeeesaaaeeseeseeeeessieeessaeeseaees 15
`Estimating Target REQUCTIONS ....0....ceceieeertee entree ee ttne nese etneeeeetnieeeee teases tiieeeeeesieeeeetiieeeeesnieeeereaa 15
`Validating Targets with Contracts Management.............:ccccccceceeceeeeeeeeeeeeseeeeeseeeeeeeeeteueeetsneeeseneee 16
`Determining True ROL ........cccccccecccecseeceeeceeceececeaeeecaeeceneeceaeeecaeeesaeesceeeescaeeecsaeessaeeesereessueeesseenenees 16
`CONCIUSION ..... cee ce ceeeceeeceeeeececeeeeceeeeceeeeceaececaaecceceeceaeeecaeeccaeescaeeeceaeeeseaeeseseesceeesceesceeseaneesnieeesseeeess 17
`
`COPYRIGHT © 2004-2006, CIRBA ING. ALL RIGHTS RESERVED
`
`PAGE 2
`
`VMware, Inc.
`
`Exhibit1012 Page 2
`
`VMware, Inc. Exhibit 1012 Page 2
`
`

`

`CIRBA SERVER CONSOLIDATION WHITE PAPER
`
`INTRODUCTION
`
`Overthe past 20 years the IT infrastructures of most organizations have moved
`away from a reliance on centralized computing power toward distributed systems.
`While the benefits of a distributed approach are numerous and well understood,
`the associated management challenges are yet to be mastered. A significant part
`of the challenge is the sprawl that can occur over time as applications and servers
`proliferate. Decentralized control and decision making around capacity,
`provisioning of new applications and hardware combined with a perceived low
`cost of server hardware have created environments with far more processing
`capacity than is required. When costis considered on a server by server basis
`this may not be troubling, but when you consider the multiples in a large
`environment, having too many servers becomesa significant burden. Simple
`math suggests that on license redundancy alone, taking even a modest numberof
`servers out of an environment savesasignificant amount on a yearly basis.
`
`This dynamic has caused many organizations to begin asking the question:
`
`“How do we consolidate some of this capacity to drive out cost?”
`
`The heterogeneous natureof distributed configurations makesthis question
`extremely difficult and time consuming to answer.
`
`This paperlooksat the challenge of IT asset optimization, specifically server
`consolidation, and how an approachthat relies on quantitative and statistical
`analysis can greatly assist in meeting the challenge.
`
`IT ASSET OPTIMIZATION DEFINED
`
`True asset optimization is multi-faceted. When considering the optimization of
`assets you must include four sub activities:
`
`RATIOMALIZ ATION
`
`Rationalization is the systematic process of removing “slack” from the system and
`re-negotiating contracts and expenditures to match actual resource needs. This is
`especially important in areas such as softwarelicensing; where being over-
`licensed has adverse budgetary impact and being under-licensed has adverse
`governanceimplications.
`
`CIP TIMIZATIORN
`
`The complementto rationalization, optimization, is the processof altering actual
`resource requirements to allow gains to be realizedin efficiency and economiesof
`scale.
`It is typically used to free up resources so they can be utilized for other
`
`COPYRIGHT © 2004-2006, CIRBA INC. ALL RIGHTS RESERVED
`
`PAGE 3
`
`VMware, Inc.
`
`Exhibit1012 Page 3
`
`VMware, Inc. Exhibit 1012 Page 3
`
`

`

`CIRBA SERVER CONSOLIDATION WHITE PAPER
`
`purposes, and can be exploited in initiatives such as consolidation and
`virtualization.
`
`TRON SOLID ATION
`
`Consolidation refers to the process of combining applications and/or data on
`systems in order to achieve economies of scale and increase aggregate resource
`utilization. Variants on consolidation include physical consolidation (merging
`similar / identical platforms into larger ones) and application consolidation
`(“stacking” multiple applications, or multiple instances of the same application, on
`a single server).
`
`VIRTUALIZATHOR
`
`Virtualization is the process of combining several OS imagesinto a single
`virtualized platform, providing economiesof scale in resourceutilization while
`maintaining a partition between operational environments. This is often used as a
`mechanism to perform consolidation, but also complicates the processof
`rationalization, and is therefore best performedin a planned and controlled
`manner.
`
`TRADITIONAL APPROACHES TO ASSET OPTIMIZATION
`
`Asset sprawl and the potential of consolidating are an increasingly populartopic in
`IT and finance circles. There are significant savings to be had through
`successfully culling surplus licenses, maintenance and hardware from
`infrastructure. The issue is that identifying the low hangingfruit is very difficult as
`the numberof parameters and variables the come into play when considering
`something like server consolidation is massive.
`
`INFORMATION RBEGHIIRED Is 4 SERVER CON SOLIDATHOM INITIATIVE
`
`Server consolidation is a complex undertaking that requires a detailed knowledge
`of the static and dynamicattributes of an environment.
`In order to accurately
`identify consolidation candidates and assurethat any plannedtransitions will be
`problem-free, a significant amount of information on the target systems is
`required.
`Information required in a typical consolidation analysis includes:
`
`e Hardware Inventory & Configuration
`»
`System models
`CPU architectures
`Non-volatile (EEPROM) settings
`Device settings
`Serial numbers
`e Operating System Settings & Files
`"OS versions and rev levels
`Kernel parameters & Registry settings
`Name service parameters
`Locale and Time zone settings
`Scheduled job configurations
`Library versions
`Local user accounts
`Installed Patches/Hotfixes
`«Security patches
`«
`Infrastructure software patches
`»
`Patch application frequencies
`
`e
`
`COPYRIGHT © 2004-2006, CIRBA ING. ALL RIGHTS RESERVED
`
`PAGE 4
`
`VMware, Inc.
`
`Exhibit1012 Page 4
`
`VMware, Inc. Exhibit 1012 Page 4
`
`

`

`CIRBA SERVER CONSOLIDATION WHITE PAPER
`
`e Application Inventory & Configuration
`«
`Application versions
`«
`Application configuration settings
`«
`Application usage
`e Middleware Configuration
`« Middleware versions
`»
`JVM versions in use
`="
`Heap settings
`"Class Paths
`e Database Configuration
`« Database versions
`«SGA parameters
`"
`Data dictionary
`«»
`Formatting and locale settings
`System Capacity and Utilization
`*
`CPU utilization
`Network I/O
`Per-processstatistics
`Device and resourcestatistics
`Platform benchmarks
`
`e
`
`Although daunting to some, this list is just a starting point — in order to perform a
`comprehensive analysis it is necessary to scrutinize system configurations and
`runtime behavior to a much deeperlevel. Subtleties within the configuration of
`systems and environments can havea drastic impact on the well-being of
`applications and businessservices; as in all aspects of server management, the
`devil is often in the detail.
`
`Viewedin this way, it is easy to see why server consolidation must start with a
`detailed understanding of the environment, a fact that must not be overlooked
`when undertaking a consolidation initiative.
`
`METHODS COMMONLY USED FOR SERVER CONSOLIDATION
`
`In most organizations the methods employedto optimize assets are often ad-hoc
`and/or unstructured in nature. Even in cases where projects are well structured,
`the absenceofsufficient data in the analysis process often means that the
`undertaking will ultimately involve a leap of faith with no guarantee of success.
`
`Common approachesto server consolidation include:
`
`Incremental: During the course of business a consolidation opportunity
`becomes obvious, a manualanalysis is performed and decisions are
`made.
`
`Departmental: Small numbers of systems are scrutinized in a specific
`area of an organization, often with objectives that are tactical in nature
`with little support in the form of information, tools or methodologies.
`
`Monumental: A high-level initiative is identified and planned, but the lack
`of a clear methodology makesit overly manualin its execution. Such
`initiatives often have a highly variable return on investment and may
`experience creeping timeframes, mainly due to imprecise objectives, ad-
`hoc analysis and the involvement of many parties (including external
`consulting firms)
`
`COPYRIGHT © 2004-2006, CIRBA INC. ALL RIGHTS RESERVED
`
`PAGE 5
`
`VMware, Inc.
`
`Exhibit1012 Page 5
`
`VMware, Inc. Exhibit 1012 Page 5
`
`

`

`CIRBA SERVER CONSOLIDATION WHITE PAPER
`
`While each approachhasits own merits, each also suffers from limitations due to
`lack of rigor, the lack of the information or the lack of a coherent approachto
`solving the problem.
`
`A QUANTITATIVE AND STATISTICAL APPROACH TQ
`
`ASSET OPTIMIZATION
`
`A quantitative and statistical approach to asset optimization is a processthat
`considersall of the critical variables discussed earlier and results in a data driven
`and accurate view of optimization potential. This approach contains four main
`steps:
`
`1. Discovery: An in-depth understanding of what is currently deployed and
`how it is configured and operating in an environment.
`
`2. Configuration comparison: An analysis of configuration synergies
`across systems in iterative one to one comparisons to identify
`opportunities to stack applications, rationalize configurations, optimize
`resourcesor virtualize OS images. Candidates are identified due to
`similarity and consideration for the cost and effort required to make
`systems similar.
`
`3. Workload analysis: Detailed workload information to enable what-if
`workload stacking analysis on candidate systems. This process effectively
`determinesif synergistic workload capacity and patterns are present and
`can be exploited to the benefit of the organization.
`
`4. Scoring: The creation of a scorecard based on the combination of
`configuration and workload data.
`
`SERVER CONSOLIDATION ANALYSIS OVERVIEW
`
`How GONSOLIDATION ANALYSIS IS USED
`
`A server consolidation analysis is typically used in one of three ways:
`
`VALIDATION OF EXISTING PLANS
`
`Data driven analysis helps validate plans by ensuring that the source and
`destination systems are compliant from a configuration perspective and have
`sufficient capacity from a workload perspective.
`In other words, it can benefit
`initiatives that are already underwayby providing independent validation and
`verification of the intended changes.
`
`DiC TED ANALYSIS
`
`Using this method for a specific project allows the targeting of specific platform
`types, business groupsor physical locations, in order to uncover consolidation
`potential and identify viable strategies.
`In other words, it provides the analysis
`mechanism when overall goals and strategies have been defined.
`
`COPYRIGHT © 2004-2006, CIRBA ING. ALL RIGHTS RESERVED
`
`PAGE 6
`
`VMware, Inc.
`
`Exhibit1012
`
`Page6é
`
`VMware, Inc. Exhibit 1012 Page 6
`
`

`

`CIRBA SERVER CONSOLIDATION WHITE PAPER
`
`
`
`CIP PORTUNISTIM TARGET IDENTIFICATION
`
`This approach, sometimesreferred to as “scavenging”, looks across broad sets of
`systems or entire environments, letting the rulesets and algorithms do the workof
`identifying candidates in a purely empirical manner. This replaces human
`resource intensive-tasks with rule-based numbercrunching, and the resulting
`scorecards often uncover “regions of compatibility” in areas that may not have
`been previously considered.
`
`Asis implied, the strategy employed is somewhat dependent on the scope and
`overall lifecycle of a consolidation initiative.
`In practice, however, all three are
`typically leveraged in order to automate the various phases of the project and
`provide an ongoing “dashboard”of optimization potential.
`
`
`
`
`64
`
`ging.
`
`n 9° & a433-key
`hphphry
`htoztly
`hewlett
`hit-tea
`ow CURE
`
`h-greenhat Le
`
`paneLeeoe
`Cage
`
`MSA Example of opportunistic target identification
`that spansplatform and application boundaries.
`In this visual System Compliance Index (SCI)
`aintogrind ERRRN MAS
`representation, higher numbers (in green and yellow)
`: RESETSRRR SS ws SSS
`hto2tly
`indicate high compatibility, while lower numbers
`SASSR :
`y y
`btt-tea
`ey Kpies
`(orangeto red) indicate low compatibility.
`‘Ss ANN “S AG SS
`ee LeeZe
`ay
`eeneywyop
`on
`hewlett
`ee
`GHEE
`RERRR
`ee
`oo
`hphphry
`ey
`This example shows high compatibility among AIX
`ee
`TRRN CA
`Uey
`and HPUX systems(in the top left and center areas,
`ee
`be ey
`SS
`“~
`SS
`WAS
`oe
`oc
`y
`oe:
`SASS
`iGEe
`respectively), and the relatively low compatibility
`SSS
`weSSN NS RS
`BA
`ie yy
`
`SAENS
`SSS: SRS
`so
`among Solaris systems (bottom right). The ruleset
`& ESS usedfor this analysis includes, among otherthings,
`— —
`AN SNS
`RSS checksfor binary andlibrary compatibility, explaining
`=
`cS SSS
`ESAS the high intra-platform and low inter-platform scores
`oe
`Ca
`SIA gh
`intra-p
`p
`=
`RAG ANSS
`
`SSSR
`tregreenhat RRRSR
`expression
`‘ken
`
`How GONSOLIDATION ANALYSIS WORKS
`
`As discussed, Server Consolidation Analysis is a multi-step process that analyzes
`multiple areas of system configuration and operation, and combinestheseinto a
`single composite scorecard. The two general areas of analysis are Configuration
`Compliance and Workload Compatibility.
`
`(ONFISURATION DOMPLIANCE ANALYSIS
`
`Configuration Compliance Analysis begins by performing a deep N-to-N
`comparison of all systems in the scope of the analysis. This yields a comparison
`matrix that indicates the complete setof differences between each serverandall
`of its counterparts.
`In such a raw form this is not yet in a state that where
`conclusions can be drawn, asit is common for servers to have manydifferences
`from one another, particularly if they perform different functions.
`
`The next step involves the application of a compliance ruleset, or consolidation
`“cookbook”, in order to analyze the differences between servers and derive
`weighted scores of how compatible they are for a particular purpose, such as
`running applications or sharing data. This step produces a System Compliance
`Index matrix, or SCI matrix, that provides a visual representation of compatibilities
`between systems.
`
`COPYRIGHT © 2004-2006, CIRBA INC. ALL RIGHTS RESERVED
`
`PAGE 7
`
`VMware, Inc.
`
`Exhibit1012 Page 7
`
`VMware, Inc. Exhibit 1012 Page 7
`
`

`

`CIRBA SERVER CONSOLIDATION WHITE PAPER
`
`NOAACISA NES
`7
`
`Woo-equlo"julyoxle
`
`
`
`woseqla"JUul'ze-Egxle
`
`[oe
`
`a433-key.int.cirba.com
`
`aix.int.cirba.com
`
`aix53-32.int.cirba.com
`
`aix53-64.int.cirba.com
`I
`[sationcom| a
`.<
`aix64.int.cirba.com
`coffey.int.cirba.com
`
`sles8.int.cirba.com
`intuition.int.cirba.com \
`messier.int.cirba.com
`
`th73.int.cirba.com
`
`fairy.int.cirba.com
`m
`s8-doc.int.cirba.co
`
`SSA
`
`-
`
`VL
`
`o
`
`~Le
`
` —~ o
`
`~
`
`Another important aspect of the configuration compliance analysis step is the
`modeling of remediation costs.
`Individual rules within a compliance ruleset may
`optionally contain a remediation cost estimate, andif that rule is not satisfied then
`the cost is factored in. This allows the analysis to provide more thanjust a
`representation of compatibility; is can also compute an aggregate remediation
`cost that represents what it would take to make the systems compliant.
`
`
`
`
`
`
`erediadieGor)
`
`
`
`
`
`
`
`
`
`
`absent
`
`
`
`
`aiser
`
`
`absart
`
`
`
`
`Ganeric_1
`$108
`
`
`aselfecovery
`
`82508
`
`
`
`
`
`
`
`
`This is useful for several types of analysis, including estimating upgrade costs and
`calculating the software overheadofvirtualization, but in consolidation analysis it
`is used to representthe costof “conditioning” a platform to allow application
`stacking.
`
`COPYRIGHT © 2004-2006, CIRBA ING. ALL RIGHTS RESERVED
`
`PAGE 8
`
`VMware, Inc.
`
`Exhibit1012 Page 8
`
`sundin.int.cirba.com
`"yul“x1eeaAaNgegaiywore|eLoeeFeFePeEUwoo'equiajulAey-eereeeggELLELLLLEEEEE
`mousuwalelele6)somoneFSehhLLrrCSs—s—SOCN
`MNquiswrhayeeeUex
`LEDe7WooeqilD"JUI"gsa|sai.GEEyeie2SYeoerrstrt—‘é=‘=FE(EEE~
`
` BqIID"}UIaISSalWEAEoe>tmCWOODBqQIID"JULUOIVNIUIwgwgtaLiLeyLGeeSs——hrmyGEEEGLy
`
`WOOD
`WEEE.LDWoo-equla"jul90p-gspadrassriedze
`woorequorwUlAstey_2-_.yaoe
`wonwtoreor178)8
`
`-_Cee7CULLawosequa
`
`eeseePees)~—hCrrlCrCULC
`
`woo-equja"jurulpuns
`
`_
`
`-
`
`a5
`\“
`
`7:
`
`XN
`\
`SX
`MS
`SSS
`SSS
`\
`N
`WN
`s
`RON
`Ss
`
`-e
`
`[os
`
`S
`
`AS
`SS
`N
`
`RSS
`ANG
`ASAS
`ASS
`~~SS
`AC
`‘
`iS
`\
`
`RS~~
`NS
`SSS
`TE
`AX
`[at .
`S
`S&S
`
`VMware, Inc. Exhibit 1012 Page 8
`
`
`
`
`

`

`CIRBA SERVER CONSOLIDATION WHITE PAPER
`
`se Mya(1A SS
`“os
`
`WORKLOAD GOMPATIBILITY ANALYSIS
`
`While configuration compliance analyzes the compatibility between systems from
`a static perspective, workload analysis looks at the dynamics of systems to
`understand if workload levels and patterns are compatible.
`
`Workload analysis starts with the tracking of key operationalstatistics of each
`system in the scopeof the analysis, and results in historical histograms of the
`operational patterns over time. Most analysis strategies make use of a few key
`stats, such as CPU utilization and network activity, but workload tracking facilities
`can track up to 100 distinct stats on a typical UNIX system, providing considerable
`breadth and strategy-specific detail to the analysis process.
`
`Target Server - base_workload-wise
`
`100°|
`
`
`
` cpu.pctsys
`
`
`
`
`
`
`
`4 20=226 8 10 12 14 #16 #18
`
`
`
`
`
`
`
`
`
`
`
`Time of Day
`
`The resulting histograms can be then used to perform a what-if analysis by
`normalizing the histograms and stacking them on one another. Using a matrix
`correlation strategy similar to the configuration compliance analysis, the selected
`workloadstatistics can be stacked on one another for each server combination to
`identify which combinations will fit within the resourcesof the target system and
`which oneswill not. This is a multi-step process that independently analyzes
`peakloads andthird quartile (75%) loads, at both like times and worst-case times
`(simulating detrimental shifts in the processing pattern of the servers), in order to
`product a weighted score of each scenario.
`
`COPYRIGHT © 2004-2006, CIRBA INC. ALL RIGHTS RESERVED
`
`PAGE 9
`
`VMware, Inc.
`
`Exhibit1012 Page 9
`
`VMware, Inc. Exhibit 1012 Page 9
`
`

`

`CIRBA SERVER CONSOLIDATION WHITE PAPER
`
`NOAACISA NES
`Sat
`
`2£3vsMy233
`333
`
`100
`
`80
`
`60
`
`40
`20
`
`of
`
`0
`
`
`
`
`
`2
`
`4
`
`6
`
`8
`
`10
`
`12
`
`14
`
`#16
`
`18
`
`20
`
`22
`
`Time of Day
`
`Source mappedto Target (Ratio 1, -9 Hour Offset)
`
`
`
`
`
`
`
`22
`
`° 2
`
`.3
`
`°o
`
`
`o 0
`
`”~
`
`~
`pa)°
`
`a 3a°
`
`2
`
`4
`
`6
`
`8
`
`10
`
`12
`
`14
`
`#16
`
`18
`
`20
`
`22
`
`Time of Day
`
`Workload Compatibility Index (WCI) Computation Scorecard:
`
`Criteria
`
`Value
`
`Score
`
`
`
`Aggregate WCI
`
`98
`
`The output of this step is a Workload Compatibility Index matrix, or WC! matrix,
`that highlights favorable consolidation candidates from a workload perspective.
`multiple stats are being analyzed, then this step is repeated for each, and the
`resulting matrices are mathematically combined to produce a single composite
`WCI matrix.
`
`If
`
`COPYRIGHT © 2004-2006, CIRBA INC. ALL RIGHTS RESERVED
`
`PAGE 10
`
`VMware, Inc.
`
`Exhibit1012 Page 10
`
`VMware, Inc. Exhibit 1012 Page 10
`
`

`

`CIRBA SERVER CONSOLIDATION WHITE PAPER
`
`NOAACISA NES
`7
`
`sundin.int.cirba.com a433-key.int.cirba.com
`eeaERoon»[=lelalato_—
`wos'eqilo"Jul",Auyeyeseepeae
`
`woo-equioqurze-eGxre[2249
`
`
`woo"equio"yjul"uommur
`Wos"eq/!D"JUlalsseauwes
`Wos"BqiID"JUI"gsa|slee
`WOd'eqIID"JUI"P9-ESX1e
`wos'equlo"jurAayjoo
`woo"equlo"jurpoxle
`wooequioqurezys
`LeBeCUAa
`LecfiaBoBBona=Ia
`oeee!aeee
`a-@s
`saeUD as
`sles
`BYYegLoiae
`se4aoaLD
`aeco
`
`
`aa.
`oo
`a.6CUSCe
`aga
`a
`eeraea
`ee
`aeVeeLyey)a
`Ay
`SENAFye7ee
`
`wooequio"jul‘ujpunsBele
`Wos"eqJ!D"JUul"D0p-gsplaleee
`
`2W
`
`woo'equioqurxie
`
`BS
`
`oo
`
`aix.int.cirba.com
`
`aix53-32.int.cirba.com
`
`aix53-64.int.cirba.com
`
`75
`
`ra
`
`aix64.int.cirba.com
`
`coffey.int.cirba.com
`
`sles8.int.cirba.com
`
`intuition.int.cirba.com
`
`messier.int.cirba.com
`
`rh73.int.cirba.com
`
`fairy.int.cirba.com
`
`s8-doc.int.cirba.com
`
`:
`
`Eee
`
`2
`ED
`a
`4e
`
`ca
`
`GENERATING 4A DOMPOSITE SOORECARID
`
`After the completion of the configuration compliance and workload compatibility
`analyses, the two can beoverlaid and mathematically combined to produce a Co-
`Habitation Index matrix, or CH/ Matrix. This is the ultimate scorecard in the
`analysis process, and provides a visual representation of consolidation
`candidates.
`
`All other analysis outputs are indexed to this matrix, allowing a promising
`opportunity to be investigated by looking into the workload scores
`the individual
`workload patterns, the compliance scores, the detailed differences between
`servers, and the remediation costs.
`
`FACTORS AFFECTING RULE SETS
`
`A key element of the configuration compliance analysis is the rule set, or
`It must reflect the nature of the opportunity being
`cookbook,that is being used.
`investigated, and different rulesets exist for different purposes:
`
`ME AVYWEIGHT / BINARY APPLICATIONS
`
`The analysis of heavyweight applications is mainly concerned with binary
`compatibility and the global settings on the system that affect the application. For
`example, the following factors are typically scrutinized when determining if two
`such applications can co-exist on a single OS image:
`
`OS version
`Maintenanceand patchlevels
`kernel settings
`name service settings
`locale and time zone settings
`
`e
`
`COPYRIGHT © 2004-2006, CIRBA ING. ALL RIGHTS RESERVED
`
`PAGE 11
`
`VMware, Inc.
`
`Exhibit 1012 Page 11
`
`VMware, Inc. Exhibit 1012 Page 11
`
`

`

`CIRBA SERVER CONSOLIDATION WHITE PAPER
`
`e
`e
`
`infrastructure software versions
`library versions
`
`A typical cookbook for this purpose will contain dozens of distinct rules, and may
`also include site-specific or application-specific conditions that must be met, such
`as the existence of a specific user account, in order to produce a high score.
`
`Hava / MET APPLIC AT HOM S
`
`Because applications that run within virtual machines are somewhatisolated from
`the OS environment, the rulesets typically focus on the environmental
`componentsaffecting their operation, and not necessarily the platform
`compatibility itself. Examples of configuration areas that are scrutinized in this
`case include:
`
`JVM versions
`class paths
`name service settings
`registry settings
`file systems and shares
`queue managersettings
`webarchives
`
`As in other scenarios, rulesets may be augmented with specific conditions that
`must be met, such as the presenceof specific Java/XML resources.
`
`Oars /COnNTeny CONSOLIDATION
`
`The consolidation of data content, such as is the focus of database consolidation
`initiatives, focuses on the compatibility of the data management infrastructure.
`Examples of system properties scrutinized in this type of analysis include:
`
`RDBMS manufacturer and version
`System Global Area (SGA)settings
`Data Dictionary settings
`Date formatting settings
`Server Time zone settings
`
`Again, many otherfactors can be considered, including security patch levels and
`other data integrity-related requirements, when performing this type of analysis.
`
`FACTORS AFFECTING WORKLOAD ANALYSIS
`
`As with configuration compliance analysis, workload analysis can focus on
`specific aspects of system operation in order to portray an accurate picture of the
`post-consolidation world.
`
`WHIDH STATISTICS TO ANALYZE
`
`Almost every workload analysis begins with the CPUutilization, and the most
`common variant to useis the overall usage (i.e. the inverted idle time, rather than
`system or usertime) asit is typically reflective of the overall impact of an
`application on the system and its compute resources.
`
`COPYRIGHT © 2004-2006, CIRBA INC. ALL RIGHTS RESERVED
`
`PAGE 12
`
`VMware, Inc.
`
`Exhibit1012 Page 12
`
`VMware, Inc. Exhibit 1012 Page 12
`
`

`

`CIRBA SERVER CONSOLIDATION WHITE PAPER
`
`Applications that communicate over networks (which means virtually all
`applications) will benefit from the analysis of network utilization, as stacking such
`applications will incur an aggregate load on the networkinterfaces that should be
`understood.
`
`Otherstatistics that can be factored into the analysis include:
`
`per-process CPU utilization
`run queues
`file system I/O
`swap space and commit charge
`inter-process communication activity (e.g. semaphores)
`
`WGING RENCHMARKS TO NORMALIZE WORKLOADS
`
`Whenperforming the what-if analysis, systems arefirst normalized to reflect their
`relative power. This is critical, as it allows the modeling of scenarios where small
`servers are being consolidated onto larger ones, or where newserversare being
`introduced. For example, a heavy workload on a low end server maytranslate
`into a negligible workload on a powerful server, and vice versa. There are several
`mechanisms available for normalizing workloads, including referencing industry
`benchmarksas well as empirical benchmarking.
`
`The useof industry benchmarksallows a consolidation scenario to be analyzedin
`a mannerconsistent with the use pattern of the application being consolidated.
`For example, applications that perform integer operations will be best represented
`if the analysis uses SPECint numbers to normalize the CPU workloads. Similarly,
`floating point, |/O and Java-specific benchmarks can be utilized depending on the
`scenario.
`
`If purely data-driven analysis is desired, then empirical benchmarking can be used
`to representthe relative power of two systems. Workload tracking facilities can be
`used to provide the data for the analysis. The results of this can be used to
`estimate relative computational powerwithout referencing industry benchmarks,
`allowing the analysis to be completely data-driven.
`
`For even more accurate results, the benchmark can be replaced with an
`application-specific benchmark that more precisely mimics the operation of the
`target application, providing an extremely accurate answer.
`
`OTHER TECHNICAL ANALYSIS FACTORS
`
`There are a variety of other factors beyond configuration and workload that can
`greatly impact the compatibility of consolidation candidates. Consideration for the
`following should be included in any consolidation exercise:
`
`CHANGE Fis TrorRy COMPATIBILITY
`
`If the sample history gathered during the consolidation analysis includes a
`sufficient span of time then the change histories of the target systems can be
`scrutinized. The main consideration with change histories is the change tolerance
`of given applications; if the target system is frequently changed in ways that may
`disrupt the operation of an application then this may be causefor concern. Patch
`activity is a example of change wherethe tolerancesof two applications may not
`liné up, and combining them onto a single OS image may bedetrimental.
`
`COPYRIGHT © 2004-2006, CIRBA INC. ALL RIGHTS RESERVED
`
`PAGE 13
`
`VMware, Inc.
`
`Exhibit1012 Page 13
`
`VMware, Inc. Exhibit 1012 Page 13
`
`

`

`CIRBA SERVER CONSOLIDATION WHITE PAPER
`
`ACRES S CONTROL COMPATIBILITY
`
`Although this is an area often covered by the configuration compatibility analysis,
`it warrants special mention.
`If an application is designed to allow user-level
`access to a server then it may be incompatible with another application that
`assumesits serveris locked-down.
`In other words, if data access rules are
`designed in a non-shared environment then moving to a shared environment may
`compromise data integrity.
`In such cases the answer may beto go to virtualized
`partitions and not share a common OS image.
`
`NON-TECHNICAL ANALYSIS FACTORS
`
`There is more to server consolidation analysis than number crunching, and there
`are a number of non-technical considerations that must be addressed in most
`initiatives. These include:
`
`BUSINESS GROUPS
`
`Crossing business group boundaries as part of a consolidation initiative can be
`difficult, as business service boundaries are often not addressed through
`empirical analysis alone. To betruly effective this must be part of an overall
`infrastructure management strategy, and take chargeback models and other
`agreements into account in the analysis phase.
`
`(RE OGRARPHY AND Time ZONES
`
`Consolidation across time zones can be problematic, as applications may be
`sensitive to the global locale settings of the system and therefore may experience
`problems if they were to change. Given this, analysis should avoid stacking
`servers that are in different locations unlessit is part of the overall strategy, and if
`itis part of the strategy then all location-specific aspects of system operation
`should be factored into the analysis.
`
`SuUCMESS FUL IMPLEMENTATION
`
`It is often helpful to combine risky changes with features or improvements that are
`considered value-add to the end user. By bundling performanceor functional
`improvementswith the overall technology transition, an implicit buy-in can be
`fostered in the user community, thus mitigating the risks of operational issues by
`providing a perceived upside thatjustifies this risk.
`
`BENEFITS OF A QUANTITATIVE AND STATISTICAL
`
`APPROACH
`
`A methodical, analytical approach to Asset Optimization provides several key
`benefits:
`
`LITTLE OR NO RESEAREH OVERHEAD
`
`Empirical and data driven results are produced from actual system data,
`signatures and rulesets rather than manual evaluation and comparison.
`
`COPYRIGHT © 2004-2006, CIRBA ING. ALL RIGHTS RESERVED
`
`PAGE 14
`
`VMware, Inc.
`
`Exhibit1012 Page 14
`
`VMware, Inc. Exhibit 1012 Page 14
`
`

`

`CIRBA SERVER CONSOLIDATION WHITE PAPER
`
`ACMMURAGTE RESULTS
`
`PeNA retin
`NOAACISA NES
`noe
`
`Becauseresults are data-driven they are not as prone to errors in judgment or
`data entry. Detailed mathematical analysis ensures that the problem is treated
`with the same degree of sophistication as the systems being analyzed.
`
`Broa, MULTIDIMENSIONAL ROVERAGE
`
`The unique combination of configuration and workload analysis give a broader
`view of opportunities and provides a higher degree of validation for existing plans.
`Furthermore, the ability to use multiple workload statistics, weighted through a
`variety of industry and empirical benchmarks, adds further dimensions to the
`overall scoring mechanism.
`
`FPLEXMIBLE RULES ETS AND BCORING
`
`Because there are severalflavors of optimization and consolidation, and because
`applications, systems and environments must be analyzed within a broader
`ecosystem, the ability to employ multiple distinct rulesets allows meaningful
`analysis to be performed and multiple strategies to be simultaneously pursued.
`
`REALIZATION OF GOST REDUCTIONS
`
`ESTIMATING TARGET REDUCTIONS
`
`The detailed hardware, OS and software inventories that are generated as part of
`the data gathering processare indispensable in the planning stages of a
`consolidation project. Accurate reporting of “actuals” at the outset of an analysis
`helps set goals and identifies areas for directed analysis.
`
`By using these inventories, a calculator can be usedto plan overall strategies and
`estimate reduction levels:
`
`CurrentNumber Target Reduction Target Number
`40%
`
`hg
`
`haDe]
`fu
`feBepyoddehooo
`
`ai4 4 53 5 -
`
`SHaaIg
`
`COPYRIGHT © 2004-2006, CIRBA INC. ALL RIGHTS RESERVED
`
`PAGE 15
`
`VMware, Inc.
`
`Exhibit1012 Page 15
`
`VMware, Inc. Exhibit 1012 Page 15
`
`

`

`CIRBA SERVER CONSOLIDATION WHITE PAPER
`
`NOAACISA NES
`
`VALIDATING TARGETS WITH GONTRACTS MANAGEMENT
`
`It is critical that any consolidation initiative target gains that are attainable. The
`act of consolidating servers is as mucha political and contractual exerciseasit is
`a technical one. To that end, it is critical that all affected parties be engaged, and
`that buy-in is sought in order to facilitate the project and mitigate risk.
`
`One very important aspectof this processis the “contracts heads-up”, a process
`in which contracts groups are consulted regarding all planned asset reductions.
`Communicating with contracts groups early on allows them to verify whetherthe
`targeted gains can be realized from a contracts perspective