2008-FAST-Avoiding the Disk Bottleneck in the Data Domain De(9)

FAST有关论文。。

Figure 4: Logical/Physical Capacities at Data Center AFigure4:Logical/PhysicalCapacitiesatDataCenterA

Figure 5: CompressionFigure5:CompressionRRatios at Data Center AatiosatDataCenterA

MinMin

DailyglobalDaily global compressioncompressionDailylocalDaily local compressioncompression

10.0510.051.581.58

MaxMax74.3174.311.971.97

AverageAverage40.6340.631.781.78

Standard Standard

deviationdeviation13.7313.730.090.09

onnewsegments),cumulativeglobalcompressionratioon new segments), cumulative global compression ratio

(thecumulativeratioofdatareductionduetoduplicate(the cumulative ratio of data reduction due to duplicate segmentelimination),andcumulativetotalcompressionsegment elimination), and cumulative total compression ratio(thecumulativeratioofdatareductionduetoratio (the cumulative ratio of data reduction due to duplicatesegmenteliminationandZiv-LLempelstyleduplicate segment elimination and Ziv-Lempel style compressiononnewsegments)pression on new segments) over time.

sttAttheendof31s day, cumulative global compression day,cumulativeglobalcompressionAt the end of 31ratioreaches22.53to1,andcumulativetotalratio reaches 22.53 to 1, and cumulative total pression ratio reaches 38.54 to 1.

Table1:STable 1:Statistics on Daily GlobaltatisticsonDailyGlobalaand Daily Local ndDailyLocal

Compression Ratios at Data Center ACompressionRatiosatDataCenterA

DatacenterAbacksupstructureddatabasedataovertheData center A backs up structured database data over the courseof31daysduringtheinitialdeploymentofacourse of 31 days during the initial deployment of a deduplicationsystem.Thebackuppolicyistododailydeduplication system. The backup policy is to do daily fullbackups,whereeachfullbackupproducesover600full backups, where each full backup produces over 600 GBatsteadystate.Therearetwoexceptions:GB at steady state. There are two exceptions:

h

Duringtheinitialseedingphase(until6tth day in this dayinthisDuring the initial seeding phase (until 6example),differentdataordifferenttypesofdataareexample), different data or different types of data are rolledintothebackupset,asbackupadministratorsrolled into the backup set, as backup administrators figureouthowtheywanttousethededuplicationfigure out how they want to use the deduplication system.Alowrateofduplicatesegmentsystem. A low rate of duplicate segment identificationandeliminationistypicallyassociatedidentification and elimination is typically associated withtheseedingphase.with the seeding phase.

hTherearecertaindays(18tth day in this example) dayinthisexample)There are certain days (18whennobackupisgenerated.when no backup is generated.

ThedailyglobalcompressionratioschangequiteabitThe daily global compression ratios change quite a bit

overtime,whereasthedailylocalcompressionratiosareover time, whereas the daily local compression ratios are quitestable.Table1summarizestheminimum,quite stable. Table 1 summarizes the minimum, maximum,average,andstandarddeviationofbothdailymaximum, average, and standard deviation of both daily globalanddailylocalcompressionratios,excludingglobal and daily local compression ratios, excluding

h

seeding(thefirst6)daysandnobackup(18tth)day. day.seeding (the first 6) days and no backup (18

Data center B backs up a mixture of structured database DatacenterBbacksupamixtureofstructureddatabaseand unstructured file system data over the course of 48 andunstructuredfilesystemdataoverthecourseof48days during the initial deployment of a deduplication daysduringtheinitialdeploymentofadeduplicationsystem using both full and incremental backups. Similar systemusingbothfullandincrementalbackups.Similar

h

to that in data center A, seeding lasts until the 6tothatindatacenterA,seedinglastsuntilthe6tth day, day,

thh

andthereareafewdayswithoutbackups(8,12-1412-14tth,and there are a few days without backups (8thth

35 days). Outside these days, the maximum daily days).Outsidethesedays,themaximumdaily35

logicalbackupsizeisabout2.1TB,andthesmallestsizelogical backup size is about 2.1 TB, and the smallest size isabout50GB.is about 50 GB.

Figure6showsthelogicalcapacityandthephysicalFigure 6 shows the logical capacity and the physical capacityofthesystemovertimeatdatacenterB.capacity of the system over time at data center B.

hAt the end of 48Attheendof48tth day, the logical capacity reaches about day,thelogicalcapacityreachesabout41.4TB,andthecorrespondingphysicalcapacityis41.4 TB, and the corresponding physical capacity is about 3.0 TB. The total compression ratio is 13.71 to 1. about3.0TB.Thetotalcompressionratiois13.71to1.

Figure 4 shows the logical capacity (the amount of data Figure4showsthelogicalcapacity(theamountofdata

fromuserorbackupapplicationperspective)andthefrom user or backup application perspective) and the physicalcapacity(theamountofdatastoredindiskphysical capacity (the amount of data stored in disk media)ofthesystemovertimeatdatacenterA.media) of the system over time at data center A.

sttAttheendof31s day, the data center has backed up day,thedatacenterhasbackedupAt the end of 31

about16.9TB,andthecorrespondingphysicalcapacityabout 16.9 TB, and the corresponding physical capacity is less than 440 GB, reaching a total compression ratio of islessthan440GB,reachingatotalcompressionratioof38.54to1.38.54 to 1.

Figure 5 shows daily global compression ratio (the daily Figure5showsdailyglobalcompressionratio(thedailyrate of data reduction due to duplicate segment rateofdatareductionduetoduplicatesegmentelimination),dailylocalcompressionratio(thedailyrateelimination), daily …… 此处隐藏：4272字，全部文档内容请下载后查看。喜欢就下载吧 ……