操作系统第七版第十二章

Chapter12:Mass-StorageSystems12.2Silberschatz,GalvinandGagne©2005OperatingSystemConcepts–7thEdition,Jan1,2005Chapter12:Mass-StorageSystemsOverviewofMassStorageStructureDiskStructureDiskAttachmentDiskSchedulingDiskManagementSwap-SpaceManagementRAIDStructureDiskAttachmentStable-StorageImplementationTertiaryStorageDevicesOperatingSystemIssuesPerformanceIssues12.3Silberschatz,GalvinandGagne©2005OperatingSystemConcepts–7thEdition,Jan1,2005ObjectivesDescribethephysicalstructureofsecondaryandtertiarystoragedevicesandtheresultingeffectsontheusesofthedevicesExplaintheperformancecharacteristicsofmass-storagedevicesDiscussoperating-systemservicesprovidedformassstorage,includingRAIDandHSM12.4Silberschatz,GalvinandGagne©2005OperatingSystemConcepts–7thEdition,Jan1,200512.1OverviewofMassStorageStructureMagneticdisksprovidebulkofsecondarystorageofmoderncomputersDrivesrotateat60to10,000timespersecondTransferrateisrateatwhichdataflowbetweendriveandcomputerPositioningtime(random-accesstime)istimetomovediskarmtodesiredcylinder(seektime)andtimefordesiredsectortorotateunderthediskhead(rotationallatency)HeadcrashresultsfromdiskheadmakingcontactwiththedisksurfaceThat’sbadDiskscanberemovableDriveattachedtocomputerviaI/ObusBussesvary,includingEIDE,ATA,SATA,USB,FibreChannel,SCSIHostcontrollerincomputerusesbustotalktodiskcontrollerbuiltintodriveorstoragearray12.5Silberschatz,GalvinandGagne©2005OperatingSystemConcepts–7thEdition,Jan1,2005Moving-headDiskMachanism12.6Silberschatz,GalvinandGagne©2005OperatingSystemConcepts–7thEdition,Jan1,2005OverviewofMassStorageStructure(Cont.)Magnetictape(Chapter12.1.2)Wasearlysecondary-storagemediumRelativelypermanentandholdslargequantitiesofdataAccesstimeslowRandomaccess~1000timesslowerthandiskMainlyusedforbackup,storageofinfrequently-useddata,transfermediumbetweensystemsKeptinspoolandwoundorrewoundpastread-writeheadOncedataunderhead,transferratescomparabletodisk20-200GBtypicalstorageCommontechnologiesare4mm,8mm,19mm,LTO-2andSDLT12.7Silberschatz,GalvinandGagne©2005OperatingSystemConcepts–7thEdition,Jan1,200512.2DiskStructureDiskdrivesareaddressedaslarge1-dimensionalarraysoflogicalblocks,wherethelogicalblockisthesmallestunitoftransfer.The1-dimensionalarrayoflogicalblocksismappedintothesectorsofthedisksequentially.Sector0isthefirstsectorofthefirsttrackontheoutermostcylinder.Mappingproceedsinorderthroughthattrack,thentherestofthetracksinthatcylinder,andthenthroughtherestofthecylindersfromoutermosttoinnermost.12.8Silberschatz,GalvinandGagne©2005OperatingSystemConcepts–7thEdition,Jan1,200512.3DiskAttachment12.3.1Host-AttachedStorageHost-attachedstorageaccessedthroughI/OportstalkingtoI/ObussesSCSIitselfisabus,upto16(15)devicesononecable,SCSIinitiatorrequestsoperationandSCSItargetsperformtasksEachtargetcanhaveupto8logicalunits(disksattachedtodevicecontroller)FCishigh-speedserialarchitectureFiberChannelArbitratedLoop-光纤通道仲裁环路可支持多达126个设备，SCSI只能在每个总线上支持15个设备。每个主适配器板带有二个FC-AL插口，很容易建立万亿字节的大容量存储;光纤通道有一个较高的性能基线，即其每个环路的最大传输速率为100Mbytes/sec,它结合通道技术和网络技术从物理层到应用层的一系列协议提供更高的带宽，更长的连接距离和更灵活的连接方式；Canbeswitchedfabricwith24-bitaddressspace–thebasisofstorageareanetworks(SANs)inwhichmanyhostsattachtomanystorageunitsCanbearbitratedloop(FC-AL)of126devices12.9Silberschatz,GalvinandGagne©2005OperatingSystemConcepts–7thEdition,Jan1,200512.3.2Network-AttachedStorageNetwork-attachedstorage(NAS)isstoragemadeavailableoveranetworkratherthanoveralocalconnection(suchasabus)NFSandCIFSarecommonprotocolsImplementedviaremoteprocedurecalls(RPCs)betweenhostandstorageNewiSCSIprotocolusesIPnetworktocarrytheSCSIprotocol12.10Silberschatz,GalvinandGagne©2005OperatingSystemConcepts–7thEdition,Jan1,200512.3.3StorageAreaNetworkCommoninlargestorageenvironments(andbecomingmorecommon)Multiplehostsattachedtomultiplestoragearrays-flexible12.11Silberschatz,GalvinandGagne©2005OperatingSystemConcepts–7thEdition,Jan1,200512.4DiskSchedulingTheoperatingsystemisresponsibleforusinghardwareefficiently—forthediskdrives,thismeanshavingafastaccesstimeanddiskbandwidth.AccesstimehastwomajorcomponentsSeektimeisthetimeforthediskaretomovetheheadstothecylindercontainingthedesiredsector.Rotationallatencyistheadditionaltimewaitingforthedisktorotatethedesiredsectortothediskhead.MinimizeseektimeSeektimeseekdistanceDiskbandwidthisthetotalnumberofbytestransferred,dividedbythetotaltimebetweenthefirstrequestforserviceandthecompletionofthelasttransfer.12.12Silberschatz,GalvinandGagne©2005OperatingSystemConcepts–7thEdition,Jan1,2005DiskScheduling(Cont.)SeveralalgorithmsexisttoscheduletheservicingofdiskI/Orequests.Weillustratethemwitharequestqueue(0-199).98,183,37,122,14,124,65,67Headpointer5312.13Silberschatz,GalvinandGagne©2005OperatingSystemConcepts–7thEdition,Jan1,200512.4.1FCFSSchedulingFCFS磁盘调度算法根据进程请求访问磁盘的的先后次序进行调度。Illustrationshowstotalheadmovementof640cylinders.12.14Silberschatz,GalvinandGagne©2005OperatingSystemConcepts–7thEdition,Jan1,2005FCFSScheduling(Cont.)Simple,intrinsicallyfair,butgenerallydoesnotprovidet