ASHRAE NY-08-002-2008 Monitoring Vibration at a Data Center《数据中心的振动监视》.pdf

1、8 2008 ASHRAEABSTRACT This paper presents an overview of typical operationalshock and vibration test parameters for datacom and coolingequipment. The test profiles can also be used as the limit or theacceptable vibration levels for data centers. The effect of thisshock and vibration to the functiona

2、l performance of the serverdepends on how the server transmits this shock and vibrationto critical components of the server. The acceptable level ofshock and vibration level information normally can beobtained from the datacom manufacturers. However, theactual level of floor vibration in a data cent

3、er is not readilyavailable. This paper is intended to provide the informationon how to record the data and what kind of data is needed toevaluate the effect of vibration to the operational or functionalperformance of the server. In addition, this paper also informsthe data center operator on what vi

4、bration data and whatmagnitude is considered normal or high. INTRODUCTIONDatacom equipment such as high performance computerservers, storage servers, networking equipment, rack-mounted equipment, and cooling equipment can induce andtransmit vibration to surrounding datacom equipment. As theserver ge

5、ts more powerful, more cooling is required whichtranslates to higher blower or pumps speed. As a result thehigher the vibration magnitude will be transmitted to the floorand to the server. In certain cases the operator of the datacomequipment can feel the vibration and express concern on theeffect o

6、f this vibration to the datacom equipment. A data center operator has multiple servers in their datacenter and expressed concerns that vibration from the coolingequipment may effect the operation of the datacom equipment.Another data center operator is concerned that the computerroom has excessive v

7、ibration from fork lift vehicles and trucksoperating outside in the warehouse which is close to the datacenter. In another location, the datacenter is on the 12th floorof a 26-story building. The next building physically attachedto the main structure has electrical power generators, whichproduce vib

8、ration and can be felt anywhere inside the build-ing. All of these effect the vibration to the datacom equipmentoperation. To evaluate the effect of these operational vibra-tions we need to know what are the limits of vibration that thedatacom equipment can tolerate, and the vibration magnitudetrans

9、mitted to the datacom equipment. The first piece of infor-mation can be obtained from the datacom equipment manu-facturers, but each datacom equipment may have differentspecifications such as a maximum g level at any frequencyrange, or certain Power Spectral Density profile. The secondpiece of infor

10、mation is the data center vibration which is mostlikely readily available. In this paper the typical test level thatdatacom equipment can be tested will be discussed. Next, thetests on datacom equipment, criteria for acceptable datacenters, and the methodology for data center measurementswill be pre

11、sented. Then some actual measurements on somedata centers will be presented. In this paper, no attempts are made to evaluate the effectof the operational vibration to the functional performance ofthe datacom equipment. The datacom equipment manufac-turer can evaluate the effect of the vibration on t

12、heir serverbased on the testing and design of the server. Typical opera-tional testing is given for guidance purposes only.TYPICAL OPERATIONALVIBRATION AND SHOCK TESTING This section will primarily focus on the shock and vibra-tion (frequency and amplitude) test levels profile for datacomMonitoring

13、Vibration at a Data CenterBudy D. Notohardjono, PhD, PE Roger R. Schmidt, PhD, PEMember ASHRAE Member ASHRAEBudy D. Notohardjono is a senior engineer and Roger R. Schmidt is an engineer at IBM Corporation, Poughkeepsie, NY.NY-08-0022008, American Society of Heating, Refrigerating and Air-Conditionin

14、g Engineers, Inc. (www.ashrae.org). Published in ASHRAE Transactions, Volume 114, Part 1. For personal use only. Additional reproduction, distribution, or transmission in either print or digital form is not permitted without ASHRAEs prior written permission.ASHRAE Transactions 9equipment. In other w

15、ords what vibration levels datacomequipment manufacturers tested their product and assured thatwhen subjected to that vibration level there will be no perfor-mance degradation of the datacom equipment. A typical transient test for a period of 30 minutes is givenin Figure 1. The vertical axis of Figu

16、re 1 is the power spectraldensity (PSD) and the horizontal axis is the frequency in Hz.The PSD is a measure of the power content with respect to thefrequency for random vibration. The PSD units are g2/Hz, andrepresent the power present in a 1-Hz-wide square filter. Thesquare root of the integration

17、of the PSD curve with respect tofrequency gives the over-all rms level of the vibration. Thisprofile represents the probability of less than 0.001 that thevibration levels at the mounting surface of the product will begreater than the levels specified in Figure 1. In general, thespecified levels app

18、ly to the three orthogonal axes except fora rack mount product where one axis is sufficient. The g rmsof this profile is 0.1 g rms. In addition, sine-on-random testtechniques are recommended to simulate the line frequency of50 Hz and 60 Hz at 0.06 g. Alternatively, 15 minute sinedwells at 50 and 60

19、Hz shall be run at 0.06 g if sine-on-randomcapability is not available. For a frequency range from 17Hz 500 Hz, the amplitude of Level 1 vibration level shown inFigure 1 is 2.2 x 10-5 g2/Hz and for Level 2, the amplitude is3.0 x 10-4 g2/Hz. For heavy rack mounted datacom equipment, Level 1 isthe app

20、ropriate test profile and for light datacom equipment,Level 2 test profile can be used. The test profile is applied tothe base or at the products support points such as the castersor leveling feet. These two levels can also be used as an upperlimit for the vibration transmitted to the data center fl

21、oor bydatacom equipment or cooling unit. A continuous vibration as shown in figure 2 is a vibra-tion which can be present in a machine throughout its productlife. A continuous vibration can produce and sustain resonance.Telecordias NEBS Telecommunication industries spec-ifies somewhat different magn

22、itudes and frequencies. TheNEBS GR63 specifies a sine sweep at 0.1g from 5-100Hz(IBM. 1992a). Occasionally, some components within the datacomequipment cabinet can have a “hot swap” capability duringwhich a component is removed and the new component isinstalled while the system is running. During th

23、is operation, ashock magnitude up to 30 g with 3 ms pulse width in the verti-cal direction and about 15 g, with 3 ms pulse width can occur(NEBS 2006). TYPICAL SHOCK AND VIBRATION MAGNITUDE RECORDED IN DATA CENTER The shock and vibration levels above were determined byconducting actual vibration meas

24、urements in datacom equip-ment raised access floors (Frey, R. 1989). The floor vibrationis the relevant vibration transmitted to the datacom equipment.The magnitudes given are an envelope of the peak amplitudes.These values can also be used as typical limits of vibrationamplitude produced by datacom

25、 or cooling equipment. There-fore, the limits can also be used as the acceptable datacomequipment room vibration levels since datacom equipment istested to meet the above limits. Actually, most datacom equip-ment can withstand higher magnitudes of operational shockand vibration. On the other hand, d

26、atacom equipment manu-facturers may not test beyond the level indicated in the previ-ous section. The solid line in Figure 3 shows the level of vibrationshown in Figure 2. The dotted line shows the threshold wherepeople start to feel the vibration (Harris, Cyril M and Piersol,Allan G. 2001) and the

27、dashed line represents a vibration levelthat is unpleasant to people. Generally, as long as people do notfeel the unpleasant vibration level, the level of datacom equip-ment room vibration is within the allowable or acceptable level.Figure 1 Typical power operational vibration test profile(IBM 1990)

28、.Figure 2 Typical operational vibration test parameters(IBM 1990).10 ASHRAE TransactionsMONITORING FLOOR VIBRATION INDATA CENTERSThis section provides an illustration of recent measure-ment of a data center floor vibration. The purpose of thiseffort is to validate and confirm the vibration levels de

29、pictedin Figures 1 and 2.Datacom manufacturers require not only the magnitudebut also the frequency related to this magnitude as key infor-mation for evaluating the effect of the vibration on the datacomequipment. The critical components of datacom equipmentperformance depends not only the magnitude

30、 but on thefrequency of the input floor vibration. A certain frequency thatcoincides with the resonance frequency of the critical compo-nent may unlatch or degrade the performance of the datacomequipment, whereas the same magnitude at differentfrequency will not affect the datacom equipment performa

31、nce.Generally, recording vibration amplitude data alonewhether it is in g-force or velocity requires lower costmeasurement instrumentation than collecting data in order toproduce magnitude and frequency content. However, it is crit-ical that g or Power Spectral Density (PSD) versus frequencydata be

32、recorded. The frequency range of the data should bebetween 5-500 Hz. To accomplish this task it may require twosets of data collection; one between 2-50Hz to capture the lowfrequency content and the second between 50-500Hz tocapture frequencies higher than 50Hz. A data center example is shown in Fig

33、ure 4. The coolingunits are shown as A/C boxes. The datacom equipment isshown as S boxes. Vibration was monitored at 9 locationsdenoted as I-1, I-2, I-3,.to III-3 In this paper only the plots of the high value are given.Figure 5 illustrates typical acceleration in g for a frequencyrange from 2-50-Hz

34、. At 22 Hz the magnitude of the vibrationis 0.0016 g which is lower than 0.07 g given in Figure 2. Figure6 shows similar information but at a frequency range from 5-500Hz. The maximum g is 0.002g at 220Hz. This value is anorder of magnitude lower than the limit given in Figure 2(0.035g). The areas n

35、ear the cooling units have higher vibra-tion than the other areas. Figure 7 and 8 show the data in adifferent format. Here, the vibration magnitude is given in thevertical axis as g2/Hz vs Hz. At 220 Hz the vertical axis is5.4x10E-5 g2/Hz which is somewhat higher than Level 1 (seeFigure 1), but this

36、 is still below Level 2 shown in Figure 1. CONCLUSION1. Level 1 (see Figure 1) is the appropriate vibration inputtest level for a heavy datacom equipment cabinet. Thislevel is intended to mimic the datacom equipment roomfloor vibration and to test that the functionality of thesystem is not affected

37、during the test. For a small and rela-tively light equipment cabinet, Level 2 can be used as atypical test level. 2. Vibration amplitudes at 0.07g are most likely the upperlimit of vibration at the base of the datacom equipment.Figure 3 Average peak acceleration of human threshold ofperception of vi

38、bration (Harris 2001).Figure 4 Data center #1.Figure 5 Floor vibration recorded at data center 1 locationII-1 from 2-50 Hz. Reproduced with permissionfrom Notohardjono (2006).ASHRAE Transactions 11This level can be considered normal data center vibrationand most likely will not induce any functional

39、 degrada-tion to datacom equipment. 3. Recent data center floor vibration measurement validatesthe limit given in Figures 1 and 2. 4. When a concern arises as to whether the data centersfloor vibration is too high, an actual measurement of datacenter vibration is warranted. Both power spectraldensit

40、y vs. frequency and g vs. frequency plots are theinformation needed to evaluate this concern. ACKNOWLEDGMENTSThe authors would like to thank Udo Jourdan, ZakidGammarra and Piet Miedema for conducting data center floorvibration measurements. REFERENCESIBM. 1990, Corporate Standard 1-9711-002, Vibrati

41、on Lev-els for IBM Hardware Products, Product Environments,Product Classes, IBM Corp. NEBS. 2006. Network Equipment Building System (NEBS)requirements: Physical Protection, Generic Require-ments GR-63-Core, Telcordia. IBM. 1992a. Corporate Standard 1-9711-007, OperationalShock Levels for IBM Hardwar

42、e Products Environ-ments, Product Classes, IBM Corp. Frey, R. 1989. Vibration Field Survey Results and IBM Stan-dards Revisions, IBM Internal Technical Report TR01.B136 Harris, Cyril M and Piersol, Allan G. 2001 Harris Shockand Vibration Handbook McGraw-Hill Professional5th edition.Notohardjono, B.

43、D. 2006. IBM Data Center Vibration Mon-itoring, IBM Internal Technical Report 2006. Figure 6 Floor vibration recorded at data center 1 locationII-1 from 5-500 Hz. Reproduced with permissionfrom Notohardjono (2006).Figure 7 Power spectral density of floor vibration recordedat data center 1 location II-1 from 2-50 Hz.Reproduced with permission from Notohardjono(2006).Figure 8 Power spectral density of floor vibration recordedat data center 1 location II-1 from 5-500 Hz.Reproduced with permission from Notohardjono(2006).