Climate protection needs energy efficient data centers - Stulz GmbH

More documents

Recommendations

Info

4 IT Hardware and Software Optimization• 4.1 OverviewThe optimization of IT hardware and software is crucialto reducing data center power consumption. Each watt ofpower saved on the IT side is one watt less which needscooling or UPS-backup. This translates into a two-foldsaving. Detailed advice on IT operations and the appropriatehard and software is beyond the scope of this guide,as its main focus is “data center design and infrastructure”.And so we can only offer a brief overview here ofthe savings potential which can be realized through IThard and software.Reduce System Power Consumption• efficient hardware is highly dependent upon productdesign• the right choice of components and software,together with precise dimensioning will reduceenergy consumptionEfficient Use of Hardware Resources• consolidation and virtualization can lead to majorreductions in energy and materials consumption• infrastructure optimization means that hardwarecan be turned off when not in useTable 2: Overview of energy saving options for servers• 4.2 ServersAs Figure 2 demonstrates, volume servers account forabout 2/3 of the energy consumed by IT hardware. Andso it follows that the volume server area harbours greatsavings potential for reducing overall IT energy consumption.This is augmented by the fact that many of today’svolume servers operate at very low capacities. Averageutilization levels of just 10 per cent are not uncommon.But low capacity operation also means poor efficiencylevels. Even when in idle mode, a server still requires considerablymore that 70 per cent of its maximum performancepower.There are two basic approaches to reducing server energyconsumption. One method is to optimize the hardwareso that it uses less electric power. The other option is toimprove the operation of this hardware so as to increasethe average utilization level of the systems. This resultsin two-fold savings: through lower power consumptionand a reduction in hardware. To achieve optimum results,both approaches should be followed simultaneously (seeTable 2).Figure 4 shows a typical distribution of the energy consumptionof individual server components. This revealsenergy savings potential: small, 2.5” hard disk drives, forexample, consume less energy than 3.5” hard drives, harddrives with lower revolution rates use less energy thanthose with high revolution rates. If hard drive access timesand data transfer rates are not critical factors in the plannedapplication, there is often no noticeable difference insystem performance.The choice of appropriate components, right-sized to fitthe individual application, has a correlative effect on theother components so that they, too, will have reducedenergy/power demands. From an energy efficiencyperspective it is preferable to deploy one large memorymodule than two smaller modules with the same overallcapacity. There are now also some highly effective,energy-efficient CPUs, fans and PSUs on the market.Server air-flow is also an area with high energy savingspotential. Servers should be designed to draw air in acrosstheir entire front surface to ensure optimum convectioncooling. Increasing the size of the air intake vent ofeach CPU socket means that the ventilation system canoperate at higher supply air temperatures or expend less10
Energy Efficiency in the Data CenterMainboard 10%Fans 13%PSUs 20%CPU 31%Memory 10%even greater when systems are used more efficiently:consolidation and virtualization are the keywords here.The term consolidation describes the process of harmonizingand amalgamating systems, applications, databasesor strategies. The aim here is usually to simplify the infrastructureand render it more flexible. The process usuallyleads to a considerable reduction in energy consumption.HDDs 16%Fig. 4: Typical breakdown of a server’s energy consumption (Source:Fujitsu Siemens Computers)energy for cooling. With the blade server form factor inparticular, it is important that the design of both ventilationsystem and power supply should be carefully coordinatedso that space is saved in the right places. Inappropriatereduction of the size of the enclosure which housesthe CPU and RAM will lead to increased heat density, willgenerate hotspots and require more energy to transportthe same volume of air through a smaller cross-sectionalarea.Considerable energy savings can thus be achieved byhardware optimization alone, and the potential is oftenVirtualization means abstraction: logical systems areabstracted from actual physical hardware. In this processresources are not dedicated but are used collectively,which gives greater flexibility as regards availability andmeans that capacities are better exploited. This can increasesystem utilization rates considerably and thus save alot of energy at the same time.Figure 5 shows an example of server consolidation andvirtualization. Overall energy consumption can be drasticallyreduced – by 50 per cent in this example – whilststill retaining the same levels of system performance andavailability, by moving 4 system workloads onto one highperformancesystem with professional virtualization. Thisis one of the more conservative examples – appreciablyhigher savings are possible, depending on the applicationin question.Server consolidation saves energyApplication 1Application 2Application 3ApplicationAPP 44Application 7Application 8Application 1Application 2Application 3Application 4Application 5Application 6Application 7Application 8System 1System 2System4Professional Virtualization10% active2KW10% active2KW10% active2KW50% active4KWTotal Energy 8KWTotal Energy 4KWFig. 5: Energy savings through consolidation and virtualization11
Page 1 and 2: Series onEnvironment& EnergyVolume
Page 3 and 4: Energy Efficiency in the Data Cente
Page 11: Energy Efficiency in the Data Cente
Page 18 and 19: spots in the racks for example, and
Page 20 and 21: Temperature levels in both the cool
Page 22 and 23: circulating air stream. In this mod
Page 24 and 25: 2.3.volume of circulated air is ver
Page 26 and 27: levels. But the appropriate air-con
Page 28 and 29: CO2 emissions caused by UPSelectric
Page 30 and 31: The requirements for the individual
Page 32 and 33: UPS systems is designed to meet the
Page 34 and 35: UPS Output Power≥ 40 to < 200 kVA
Page 36 and 37: There are two basic construction de
Page 38 and 39: 7 Energy ContractingCooling and pow
Page 40 and 41: 8 Glossary19“-CabinetBPSBrineChil
Page 42: The German Association for Informat

Climate protection needs energy efficient data centers - Stulz GmbH

Create successful ePaper yourself

Delete template?

Save as template?