17292381 Optimize the Cooling in Your Existing Data Center

8/3/2019 17292381 Optimize the Cooling in Your Existing Data Center

1/9an Networking eBook

Optimizethe

coolingin

your existingdata center


2/91 Optimize the Cooling in Your Existing Data Center, an Internet.com Networking eBook. 2009, WebMediaBrands Inc

Optimize the Cooling in Your Existing Data Center

he number o servers is causing power andcooling demands to rise in most data centers. Thisincreasing need or cooling is not only negativelyimpacting operating expenses but also constrain-ing growth because additional cooling needs cant

be met with the existing inrastructure and requires either sub-stantial capital investment in the existing acility or investmentin a new data center. What some data centers may be over-looking are optimization methods that can increase the avail-able cooling capacity substantially while also reducing costs.

This eBook will provide

an overview of:

Thecurrentpressuresacing data centersrelating to powerand cooling

Theneedtobaselinecurrentcooling utilization and costs

Commonimprovementopportunities including bothprocess and technicalconsiderations

The Need for

Improved CoolingComputing,storage,andcommunicationtechnologiesallre -quire electricity to operate. Two observed trends are increases

in the density and perormance o each. With these trends havecome increased energy demands and expenses. This is espe-ciallyevidentindatacenterswherethereareoftenhundreds,

if not thousands, ofdevices running ona continualbasisIn fact,data centers are consuming signicant amounts opower. The Department o Energy conducted a study in 2006and reported that data centers consumed 61 billion kilowathours o energy or 1.5 percent o energy used that year. Thestudy then went on to project that demand could double by2011,potentiallyrequiringthat10additionalpowergenerationacilities be built.1

Numberssuchas these,coupledwithanecdotalevidenceindicatethatenergyconsumptionwithindatacenters,while

signicant today,will becomeincreasingly problematic i leunmanaged. One o the chieconcerns is that in tandem withthe need or power comes thedemand or cooling. The various systems in a data cente

generate heat as they operateand this heat must be removedor the data center will becometoo hot and the component ailure rate will increase.

Exacerbating the power problem is that not only do the ITsystems require power but the

coolingsystemsthemselvesconsumepower.Ingeneral,onewatt o power consumed in IT requires one watt o powerequired or cooling.

There are four challenges with this scenario. First, datacentersarendingthateithertheirsiteinfrastructurecannotaccommodate more power or their utility cannot provide more

OptimizetheCoolinginYourExistingDataCenterBy George Spaord

T

energyconsumptionwithindatacenters,whilesignicanttoday,willbecomeincreasingly

problematic i let unmanaged.




power.Second,datacentersarendingthattheirfacilitiesarenotdesignedtohandleadditionalcoolingdemandsefcientlyoreffectively.Third,thecapitalandoperatingexpensesas-

sociatedwithpower andcooling are rising rapidly.Fourth,giventheeconomy,bothcapitalandoperatingexpensesbud-getsarelimitedand,inmanycases,shrinking.Thismeansthat budget growth in these areas will come at the expense oanotherareasuchasnewservicedevelopment.Inshort,thecombination creates risks that need to be managed.

To address these challenges, two things must happen:energy consumption in IT systems must be managed, ifnotreduced.Atthesametime,coolingmustbecomemoreeffectiveandefcient.ThepurposeofthiseBookwillbetodiscuss means or existing data centers to address demandsor additional cooling.

Baseline Current EnvironmentTherststepistounderstandthecurrentstate.Therearetwoaspectstounderstand:demandandsupply.Itisimportanttounderstand cause-and-eect o the improvements perormedand also to demonstrate to management and other stakehold-ers the results o improvement activities.

Fromthedemandside,thismeansunderstandingthepowergoing into the data center and how its consumed with what-ever level o granularity possible. Rack-level or ultimately com-ponent-levelconsumptiondataistheideal,butinitiallyunder-

standingtotaluseisavalidrststep.

Fromthesupplyside,organizationsareimplementingincreas -ingly sophisticated sensor grids to understand the tempera-ture and humidity in the data center along with cooling-relatedpowerconsumption,andotheroperationaldatapointsfromthe cooling systems themselves.

There is a growing body o studies and recommendationsregarding how to deploy sensors to monitor environmen-talconditionsandpower.Ingeneral,thetypesofsensors,location,anddensityshouldbedrivenbycostsandbenets.Itispossibletocollectdataatthedevicelevel,intheaisles,

airvelocityinplenums,andsoforth.Alongwiththesensorsyou need sotware tools that can help sta with the analysiso the data so they understand trends and generate variousalertlevelswhendenedconditionsaremet.

Improvement OpportunitiesThere are many ways to improve the cooling capacity o exist-ing data centers. The ollowing are common opportunities thatmightyieldsignicantimprovementinyourenvironment:

Reduce Cooling DemandThe rst step to improving available cooling capacity is toliterallyreducethedemandforcooling.Itmayseemsimple

andsomeaspectsofitare.Theveryrstthingtodoistalkto sta and discuss their ideas and observations on howto reduce the power being used by IT equipment.

Aneasyrststepthathaslowcostsandhighpotentialbenetsistoidentifyandremoveghosts.Theseareserversthatarestillrunning,consumingpowerandcoolingbuttheservicethey were providing to the business is no longer needed andnobody told IT to decommission the systems. Ghosts can beidentiedwithnetworkmonitoringandsystemsmanagementtools,whichcanidentifysystemswithlittletononetworktrafcand/orCPUutilization.

Another approach is to consolidate applications and ITservices into as ew as possible. Large organizations thathavegrownorganically,orhadextensivemergersandacquisitions,arelikelytondmanyneedlesslyredundantapplicationsandservicesbeingprovidedtothebusiness.Forexample,10 accounting packages in use versus one. The duplicatesystems should be identied and actions identied. Someduplicateswillbeveryeasytoconsolidate,whereasothersmay take substantially more eort to gain approval.

Reducing the number o servers and volume o data storagethrough consolidation and virtualization is another method

The whole intent is to reduce the amount o power being consumed and thus heat being generated by IT equipment thamust be cooled.

Whereas the above options all mention servers, it isimportant to note that o the power going into the datacenter,onlyapproximately30percentisactuallyconsumedby IT systems. The majority is consumed by environmentasystems (45 percent) then power related inrastructure(24percent),andthenasmallamountforlightingroundingout the list (1 percent).2 Depending on how a data center isdesigned and the equipment used, these levels can varysignicantly and understan ding the current state can help

prioritize improvement approaches. One can imagine therewill be data centers that nd large savings outside oservers. For example, some organizations may nd thathey can recover a signicant amount of cooling bymoving all of the UPSs out of the climate-controlleddata center.

A last approach to mention is to evaluate i some level o outsourcing to collocation and cloud computing vendors may




help alleviate resource constraints and reduce costs. Thereisalot toconsiderbeforepursuingthisapproach,includingtechnicalarchitecture,legacysystemsupport,informationse-

curity, andregulatorycompliance tonamea few.However,cloud computing is gaining a strong ollowing and even busi-ness leaders are taking note the savings are out there andthe risks o using a cloud vendor may actually turn out to bequite low.

Increase Data Center TemperaturesNow we will shit our attention to the environmental systems.Someorganizationsmayndthattheycanincreasethetem -peratures in their data centers and immediately reduce cool-ingdemands.TheAmericanSocietyofHeating,Refrigeration,andAir-conditioningEngineers(ASHRAE)hasincreaseditsrecommended temperature to cooling and humidity require-ments to 18 to 27C (68 to77F) dry bulb temperatureand 40 to 55 percent relativehumidity.3

Despite ASHRAEs recom-mendations it is still importantto review vendor recommen-dations and warranties. It maybe that some systems requirea lower temperature and run-ning outside of the speci-

cation will invalidate warran-ties. Thus, someresearchisneeded and risks consideredbeore simply raising the tem-perature.

There is one critical miscon-ception to clariy at this point airtemperatureismeasuredonthecoolairintakeside,notonthe exhaust side. It is the temperature o the incoming air thatis expected to cool the systems that matters. Organizationsthat have thermostats located by exhausts where there is heat-ed air are unnecessarily cooling the environment as a result.

Improve Computer Room Air Conditioning

(CRAC) and Humidity CoordinationAnother common error is ailure to coordinate the activitiesof cooling systems and humidication systems. Traditionalhumidierspass warmairacross water to causeevapora-tion and then ans move this relatively warm humid air intothe data center to reduce the risk o static. The key hereis that warm air is used and that warm air can cause thecooling system to attempt to cool it. This then reduces the

humidity,causingaloopwherebothsystemsrununnecessarily.The air movement can be analyzed and management practices have evolved to limit this battle or supremacy between

the two systems. Another approach is to replace the oldheat-basedhumidierwithnewtechnology,suchasanultrasonicsystem,thatcanintroducewaterparticlesintotheaiwithout heat.

Utilize EconomizersEconomizers are one o the leading means to reduce coolingcostswhileincreasingcapacity.Essentially,theenvironmentis leveraged to cool air beore the chiller to reduce the amounofmechanicalcoolingrequired.Incoldenvironments,therare organizations that rely on the economizer and use verylittle mechanical cooling. The approach has earned the titleree cooling.

There are two types o economizers air side and fuidside. The air side systemuses external air when certain conditions are true. Therelatively cool external air ispassed through lters to remove particulates, humidied/dehumidiedasrequiredand then introduced into thedata center.

Fluid side systems have aliquid coolant that is cooledvia the outside air, relatively cool ground, or external water source, etcand then passed back to aheat exchanger that is used

to cool data center air. This is method considered aclosed system because external air and any potentiacontaminates it may contain arent introduced.

Depending on the geographic location of a data center

economizersmayyieldbenetsatnight,seasonally,ornearlycontinuously.Thisis denitelya technologyfor constrainedand cost-conscious data centers to investigate.

Optimize Air FlowData centers that rely on airfow or cooling need to careullyassess how air is moved and then take steps to optimize thefows o cold and hot air. There are many potential improve-mentopportunitiestoevaluateincluding:

Economizers are oneo the leading means to

reduce cooling costs while

increasing capacity.




Overhead Diffusers: air that is delivered via overheadplenums need diusers that are designed to route the air

straight down in ront o the devices that need to be cooled(assuming the air inlets are in ront). General-purpose

four-way diffusers like those seen in general-use ofceareas should be avoided due to the relative lack o ocusedair delivery.

Nonstandard Intakes and Exhausts: IT equipment with coldair intakes or exhaust ports that do not conorm with thetypical ront-to-back airfow model should be assessed andcorrective action taken. Either the equipment should be replaced with equipment that conorms to the standards othe data center or some orm o ducting or bafes shouldbe used to re-direct the intake and exhaust to meet

standards.

Face Blanks: when racks are not ully populated then theopenings should be sealed with blanking plates to preventuncontrolled airfow.

Flexible Ducting:shouldbeinspectedandanykinks, obstructions,and/orsharpbendscorrected.Junctions

should be sealed appropriately. Any tears or cuts in theducting should either be repaired or the ducting replaced.

Variable Speed Fans: should be utilized to proportionately

increase/decreasespeed(andthusenergyuse)depending on demand.

Computational Fluid Dynamics (CFD): can be used tomodel airfow in a data center and help identiy potentialproblems and opportunities or improvement.

Routine Inspections: the entire cooling system should beinspected on a scheduled basis. Various components

couldbeexaminedonarollingschedule,butitsimportantthat there are ormal inspections looking or variousproblems that require correction or opportunities toimprove.

Utilize Liquid CoolingAsanotherexampleoftheoldbecomingnew,liquidcoolingatthe rack and device level is returning. The use o water to coomainrames has been around since the dawn o electroniccomputing.Withtheintroductionofdistributedsystems,theuse o fuids declined but is now making a strong resurgenceThesimplereasonisthatliquidcoolingisfarmoreefcientthan air cooling literally thousands o times more.4

The use of the Hot Aisle/Cold Aisle Layout: this provenmethod arranges racks so that cool inlet air and warm

exhaust air are segregated. Not only does this allow or the inletairtobeascoldaspossible,butitalsoallowsfor

the exhaust air to be as hot as possible. This latter part isimportant because the hotter the exhaust air is and thegreater the dierence between that temperature and the

ambienttemperatureoftheeconomizer,thenthegreaterthe benetsofeconomizerswillbe.

Installation of Partitions: barriers are put in place toprevent the unmanaged comingling o warm and cool air.Approaches range rom installing ceilings directly abovethe cold aisle to using various materials that are hung rom

theceilingtosegregatethehotandcoldaisles,different

temperatezones,andsoforth.

Optimize Raised Floor Air Pressure and Flow: one methodo delivering cool air is to pressurize a raised foor space todeliver a determined amount o air volume through

perforatedoortiles.Toaccomplishthis,theraisedoorneeds to be o the correct height and supported bychillers and variable speed ans appropriately sized to

deliverthespeciedvolumeofcoolair.

Clean Under the Raised Floor:debris,tangledcables,andso on can accumulate under a raised foor and hamper air

fow. Routine inspection and cleaning are needed to ensureairowsasefcientlyaspossible.

Cable Management: Network and power cables that arenot properly managed can become a tangled barrier toair fow. Organizations should assess whether to locate thecables under the raised foor or overhead in trays. Theideal layout will minimize negative impacts on cooling.

Seal Openings in the Raised Floor: any place where airleaks rom the raised foor should be sealed. This includes

cableopenings,seams,andtheperimeteroftheoor.Air leakagecansignicantlyimpacttheefciencyand

eectiveness o the raised foor.

Perforated Floor Tiles: must be in the correct locations alignedwiththeairintakes,whichareoftenatthefrontof racks.Furthermore,tilesthatarenolongerneededshould

be removed and replaced with solid tiles to conserve preciouscoldair.Personnelneedtounderstandhowthe tilesaretobeusedandtheirmovement/installation/

removal should be governed by the enterprise ITchange management process.




Liquid systems can be installed in existing data centers toprecisely address the cooling needs o the new high-peror-mance high-thermal density blade systems. Liquid cooling

technologies are evolving rapidly in response to needs andoer proven cost-eective solutions.

Improve Facility InsulationOne basic element is to ensure that the data center is properlyinsulated to minimize the infuence o changes in environmen-taltemperaturesduringthecourseoftheday.Externalwalls,ceilings,andsoonallneedtobeappropriatelydesignedtoprotect the data centers environment.

Understand the OpEx vs. CapEx Trade-OffSome data centers are quite old and their cooling inrastruc-

ture dates back to the construction o the building. Theseoldersystemsarebuiltonoldertechnologies,differentdesignconcepts,andsoon.Abroadgeneralstatementthatcanbemadeisthatnewsystemsaremoreefcient.Organizationsthat have a culture o run it until it breaks risk having higheroperating expenses due to the age o the systems. Sometimesthisisduetophilosophiesaroundtheuseofcapital,andman -agement needs to understand that there are situations wherecapital investment can result not only in additional coolingcapacity, but also in lower energy consumption from thecooling systems. In short, opportunitiesmayexistwherebyinvesting capital operating expenses can be reduced,sometimes dramatically.

Move Towards Less Environmentally

Sensitive IT HardwareThe long-term direction o the industry must be toward thedevelopment and use o systems that are more tolerant ohigher-temperatures and variability in the overall environment.Continuedpressureappliedbycustomers(datacenters)willcompel vendors to continue to improve their systems to dothis.MicrosoftsFourthGenerationDataCenterprogramisan example o a data center vision moving in this direction.5

Foster Teamwork: Data Center Facilities

and IT Engineering GroupsIn some organizations there is a barrier between the en-gineering group responsible or the data center acility andtheITengineeringgroupresponsibleforITservers,storage,networking,andsoon.Thisschismneedstobeclosedandthe groups must work together to better address demandsor power and cooling. This may mean changes in reportingstructure,compensation,andsoon.Theexactmethodwillvary depending on the organization however the messagemust be very clear these two groups must work together toimprove capacity and lower costs.

rganizations are very concernedabout the environment for avariety of reasons these days.For IT, one of the challenges isthat our services fundamentally

need energy to operate. Literally, if weturn off the power then we shut down allthe IT services. In the current socioeco-nomic situation of skyrocketing energy

costs, concerns over global warming, andthe need to manage risks, IT organiza-tions would be very well served to take a

closer look at the ITIL Capacity Manage-ment process as a means to bettermanage energy consumption.The idea behind the Capacity Manage-ment process is to provide computingresources to the organization in amanner that makes the business sense.The premise is very straightforward:

through advance planning andmonitoring of the current state (in

terms of the business, IT services andIT systems), proactive decisions can bemade that lead to better procurementand/or development decisions. In turn,this reduces reactive decision-makinthat is inevitably rushed and results insuboptimal outcomes including the poten-tial for higher costs and lower quality.In terms of energy, it can be viewed thesame as any other resource with limita-tions. There are limits to the power

available from the utility today andlimits as to what can actually get intothe data center. There are also limits togrowth as some groups are nding out.

Their electric utilities are telling themthat additional power is not availablefrom the grid they are on.

O

Capacity Managementand Green IT

continued




Understand the Importance of ProcessesFormal processes that are designed and implementedcorrectly can help ensure that there is adequate cooling

capacity available.

Three important processes to consider are:CapacityManagement:theimpactsofnewITservicesor

changed IT services on power and cooling need to be understoodbeforetheygointoproduction.Capacity

Management is an ITIL process that can do just that.

ChangeManagement:Allchangesinproductionrepresent risk.Additionalserversaddedtoadatacenter,oortiles thataremoved,andcoolingductsthatarereroutedcould

all impact the cooling o critical systems in done incorrecly.

AChangeManagementprocesscanhelpassesandproperly manage risks.6

ITAssetManagement:Thisisadisciplinethatcanhelp ensurestandardsarefollowedduringprocurement, thatassetlifecyclesareunderstood,thatthetrade-offs betweenoperatingandcapitalexpensesaremanaged,

and so on.

Pursue Continuous Improvement:

Plan-Do-Check-Act (PDCA)Likeeverythingelse,coolingwillneedtoevolveovertimetomeet demands. Organizations must adopt a pragmatic stance

o planning to meet demands. Depending on the level o workrequired, projectsmay be needed or perhaps task ordersor simpler actions but they should always work through theChangeManagementprocesstobetterensurethatrisksareproperly mentioned.

The next part is to check the various processes and the stateo cooling in the data center. It is critical that objective databe collected about the environment and the ability tounderstand the causal eects between changes in processesand/ortechnologiesandtheresultingimpactstothecoolingcapacity o the data center. This need cannot be stressed

enough.Inadditiontomonitoring,scheduledreviewsmustalso take place to review the current state o processesand technologies to determine what opportunities orimprovement may exist.

Given the results o the check phase are to actually takecorrective actions and then the loop returns to planning to de-termine what to do next to best support the cooling needs othe data center and how it will support the overall business.

Where It GoesAnother challenge to factor in is the ITequipment alone doesnt account for allthe power a data center uses. In fact, itis common to see the IT equipment only

account for 30 percent of the power con-sumed. The other 70 percent goes to cool-ing, power infrastructure, lighting, etc.So when power is being planned for, theoverall needs of the system in which theIT equipment resides must be taken intoaccount as well.

For each watt of IT equipment added in adata center, there also will be one incre-mental watt of cooling needed. Thereneeds to be sufcient cooling capacity to

keep IT equipment within dened temper-ature specications to avoid heat related

incidents. With todays dense, power in-tensive systems maintaining an accept-able temperature can be a challenge andpresents another resource constraint tomanage because cooling is itself limited

and it also requires electricity.There is another important reason tomanage capacity as well. In general, as

the level of utilization of electrical sys-tems goes down, so do their efciencies.

Thus, a UPS running at 50 percent willbe less efcient than a UPS running at

90 percent of stated capacity. This is alsotrue for power supplies, cooling and othersystems. Traditional approaches of oversizing various systems to play it safe

can result in higher energy costs andneed to be replaced with more deliberatemanagement approaches.

continued




Moving Ahead From HereHopefully this document has given you some ideas toinvestigate. Be sure to move ahead with careul planning.

Understand the current state and have baseline data.Introduce changes and analyze the results take the time tounderstand the causality o the actions undertaken. Did myhoped or results materialize? and i they didnt then under-stand why and change your approach. The idea should beto develop a roadmap or improving cooling capacity. Followthe roadmap and make adjustment as necessary such thatrisks are minimized and the value created maximized.

Additional ResourcesThe ollowing are Web pages where additional inormation

canbefound:

TheNewse-mailnewsletterispublishedweeklybytheauthor o this eBook and covers many aspects o ITmanagement including Green IT by providing links to

newsstoriesandresources: www.spaffordconsulting.com/dailynews.html

Newresourcesareappearingallthetimeandacollection ofWebsitesofinterestcanbefoundat: www.spaffordconsulting.com/GreenITResources.html

TheGovernanceofGreenITisabookby

George Spaord that covers governance andprocess issues around power and cooling in the

datacenter: www.itgovernance.co.uk/products/2106

Internet.comWebcastsonGreenITinclude: LeveragingSoftwaretoImproveEnergyEfciency http://solutions.internet.com/5346_default

Governing IT in a Green World http://solutions.internet.com/5341_default ImplementingaGreenDataCenter http://solutions.internet.com/4991_default

Endnotes1ReporttoCongressonServer andDataCenter EnergyEfciency.PublicLaw109-431.U.S.EnvironmentalProtec-tionAgencyENERGYSTARProgram.August2,2007.http://www.energystar.gov/ia/partners/prod_development/downloads/EPA_Datacenter_Report_Congress_Final1.pdf

In response to concerns over energyconsumption, groups need to review andformalize their Capacity Managementprocess. At a policy level there needs tobe guidelines and standards set forthabout the organizations direction forgreen IT, and expectations around en-ergy consumption and then the processdesigned and implemented accordingly.

At a process level, there needs to beintegration with project management,procurement and Change Management

to ensure that current capacity is under-stood for electricity, cooling, and so forthand that the potential impacts of new orchanged services are clearly understoodboth at the point of initial implementationand trended over time.

Part of this necessitates that Capacity

Management have the appropriate toolsto monitor and model the various re-sources that it is responsible for. This in-cludes access to relevant electrical bills,

tools for monitoring power consumptionto the data center, or even the rack anddevice levels. It also includes understand-ing relationships between business activ-ities, the supporting IT services, and theresulting power demands.

Monitoring current consumption levelsand tracking trends are critical. Throughanalysis of the variance between plannedand actual, organizations can under-stand how they are progressing and use

this as a feedback loop for continuousimprovement and future budgets.Reporting needs to be developed that pro-vides the necessary management infor-mation for command and control to therelevant stakeholders. There are various

metrics being developed in the industry

continued




2GuidelinesforEnergyEfcientDataCenters.TheGreenGrid.2007.http://www.thegreengrid.org/

3ASHRAE,2008ASHRAEEnvironmentalGuidelinesforDatacomEquipment,(ASHRAE,2008),availableafterregistrationatwww.ashrae.org/publications/page/1900

4DonReisinger.KeepCoolwithLiquidCooling.Processor.com.August1,2008.http://www.processor.com/editorial/article.asp?article=articlespercent2Fp3031percent2F22p31percent2F22p31percent2F22p31.asp

5 Mike Manos Our Vision or Generation 4 Modular DataCentersOneway ofGetting it justright,LooseBolts,BlogPosted2December2008.http://loosebolts.wordpress.com/2008/12/02/our-vision-for-generation-4-modular-data-

centers-one-way-o-getting-it-just-right

6 The IT Inrastructure Library (ITIL) provides very goodguidance on how to design many processes. It is still up to anorganization to look at these reerence processes and decidehowbesttoimplementaprocessgiventhecurrentculture,stateofprocessmaturity,availabletime,supportingtechnology and so on.

George Spaord is a Principal Consultant with Pepperweedand an experienced practitioner in business and IT operations.He is a prolifc author and speaker, and has consulted and

conducted training on regulatory compliance, IT Governance,and process improvement in the U.S., Australia, New Zealandand China. Publications include: The Governance o Green IT,Greening the Data Center and co-authorship o The Visible OpsHandbook and Visible Ops Security.

George holds an MBA rom Notre Dame, a BA inMaterials and Logistics Management rom Michigan StateUniversity and an honorary degree rom Konan Daigaku inJapan. He is an ITIL Service Manager, TOCICO Jonah and aCertifed Inormation Systems Auditor (CISA). George is a currentmember o the IIA, ISACA, ITPI, ITSMF, and the TOCICO.

to assist with the tracking of energy con-sumption. GreenGrid.org has identied

Power Usage Effectiveness (PUE), DataCenter Effective (DCE), and short-termmetrics to track and then longer termthe Data Center Performance Efciency

(DCPE) metric. Stakeholders must beconsulted to understand what they needto know and how it should be presentedto make effective decisions.

In closing, energy costs are going up andmanagement attention is focusing on how

to not only manage these costs but alsominimize impacts to the environment.Through effective and efcient Capacity

Management, organizations can achieveboth in a sustainable manner.

Documents

17292381 Optimize the Cooling in Your Existing Data Center