Acer Altos 19000PRO4 Bedienungsanleitung

DAP Spr.‘98 ©UCB 1Lecture 14: I/O Benchmarks, Busses, and Automated Data LibrariesProfessor David A. PattersonComputer Science 252Spring 1998

DAP Spr.‘98 ©UCB 10Typical File Server ArchitectureKernel NFS Protocol & File ProcessingTCP/IP Protocols Unix File SystemEthernetEthernetDriverDis

DAP Spr.‘98 ©UCB 11AUSPEX NS5000 File ServerPrimaryMemoryPrimaryMemoryHostProcessorHostMemoryEthernetProcessorFileProcessorStorageProcessorEnhancedVME

DAP Spr.‘98 ©UCB 12AUSPEX Software ArchitectureUnix System Call LayerVFS InterfaceNFS Client LFS ClientHost ProcessorLFS ClientNSF ServerProtocolsNetw

DAP Spr.‘98 ©UCB 13Berkeley RAID-II Disk Array File Server8 Port InterleavedMemory (128 MByte)8 x 8 x 32-bitCrossbarVME VME VME VMEVMEXORX-BusBoard8 P

DAP Spr.‘98 ©UCB 14CS 252 Administrivia• Email URL of initial project home page to TA?• Upcoming events in CS 25219-Mar Thu Send in Project Survey #2

DAP Spr.‘98 ©UCB 15I/O Benchmarks• For better or worse, benchmarks shape a field– Processor benchmarks classically aimed at response time for fixed si

DAP Spr.‘98 ©UCB 16I/O Benchmarks• Alternative: self-scaling benchmark; automatically and dynamically increase aspects of workload to match characteri

DAP Spr.‘98 ©UCB 17I/O Benchmarks: Transaction Processing• Transaction Processing (TP) (or On-line TP=OLTP)– Changes to a large body of shared inform

DAP Spr.‘98 ©UCB 18I/O Benchmarks: Transaction Processing• Early 1980s great interest in OLTP– Expecting demand for high TPS (e.g., ATM machines, cred

DAP Spr.‘98 ©UCB 19I/O Benchmarks: TP by Anon et. al• Proposed 3 standard tests to characterize commercial OLTP– TP1: OLTP test, DebitCredit, simulate

DAP Spr.‘98 ©UCB 2Review: A Little Queuing Theory• Queuing models assume state of equilibrium: input rate = output rate• Notation: r average number o

DAP Spr.‘98 ©UCB 20I/O Benchmarks: TP1 by Anon et. al• DebitCredit Scalability: size of account, branch, teller, history function of throughput TPS N

DAP Spr.‘98 ©UCB 21I/O Benchmarks: TP1 by Anon et. al• Problems– Often ignored the user network to terminals– Used transaction generator with no think

DAP Spr.‘98 ©UCB 22I/O Benchmarks: Old TPC Benchmarks• TPC-A: Revised version of TP1/DebitCredit – Arrivals: Random (TPC) vs. uniform (TP1)– Terminals

DAP Spr.‘98 ©UCB 23I/O Benchmarks: TPC-C Complex OLTP• Models a wholesale supplier managing orders• Order-entry conceptual model for benchmark• Worklo

DAP Spr.‘98 ©UCB 24I/O Benchmarks: TPC-D Complex Decision Support Workload• OLTP: business operation• Decision support: business analysis (historical)

DAP Spr.‘98 ©UCB 25I/O Benchmarks: TPC-W Transactional Web Benchmark• Represent any business (retail store, software distribution, airline reservation

DAP Spr.‘98 ©UCB 26TPC-C Performance tpm(c)Rank Config tpmC $/tpmC Database1 IBM RS/6000 SP (12 node x 8-way) 57,053.80 $147.40 Oracle8 8.0.4 2

DAP Spr.‘98 ©UCB 27TPC-C Price/Performance $/tpm(c)Rank Config $/tpmC tpmC Database1 Acer AcerAltos 19000Pro4 $27.25 11,072.07 M/S SQL 6.5 2 D

DAP Spr.‘98 ©UCB 28TPC-D Performance/Price 300 GBRank Config. Qppd QthD $/QphD Database1 NCR WorldMark 5150 9,260.0 3,117.0 2,172.00 Teradata 2

DAP Spr.‘98 ©UCB 29TPC-D Performance 1TBRank Config. Qppd QthD $/QphD Database 1 Sun Ultra E6000 (4 x 24-way) 12,931.9 5,850.3 1,353.00 Infomix

DAP Spr.‘98 ©UCB 3Review: Redundant Arrays of Disks (RAID) Techniques• Disk Mirroring, Shadowing (RAID 1)Each disk is fully duplicated onto its "

DAP Spr.‘98 ©UCB 30TPC-D Performance 1TBRank Config. Qppd QthD $/QphD Database 1 Sun Ultra E6000 (4 x 24-way) 12,931.9 5,850.3 1,353.00 Infomix

DAP Spr.‘98 ©UCB 31SPEC SFS/LADDIS Predecessor: NFSstones• NFSStones: synthetic benchmark that generates series of NFS requests from single client to

DAP Spr.‘98 ©UCB 32SPEC SFS/LADDIS• 1993 Attempt by NFS companies to agree on standard benchmark: Legato, Auspex, Data General, DEC, Interphase, Sun.

DAP Spr.‘98 ©UCB 33Example SPEC SFS Result: DEC Alpha• 200 MHz 21064: 8KI + 8KD + 2MB L2; 512 MB; 1 Gigaswitch• DEC OSF1 v2.0• 4 FDDI networks; 32 NFS

DAP Spr.‘98 ©UCB 34Willy• UNIX File System Benchmark that gives insight into I/O system behavior (Chen and Patterson, 1993)• Self scaling to automatic

DAP Spr.‘98 ©UCB 35Example Willy: DS 5000 Sprite UltrixAvg. Access Size 32 KB 13 KBData touched (file cache) 2MB, 15 MB 2 MBData touched (disk) 36 MB

DAP Spr.‘98 ©UCB 36Sprite's Log Structured File SystemLarge file caches effective in reducing disk readsDisk traffic likely to be dominated by wr

DAP Spr.‘98 ©UCB 37Willy: DS 5000 Number Bytes Touched• Log Structured File System: effective write cache of LFS much smaller (5-8 MB) than read cach

DAP Spr.‘98 ©UCB 38Summary: I/O Benchmarks• Scaling to track technological change• TPC: price performance as nomalizing configuration feature• Auditin

DAP Spr.‘98 ©UCB 39Review: Storage System Issues• Historical Context of Storage I/O• Secondary and Tertiary Storage Devices• Storage I/O Performance M

DAP Spr.‘98 ©UCB 4Review: RAIDRAID sales• 1993: $3.4 billion on 214,667 arrays (≈ $15,000 / RAID)• 1996 forecast: $11 billion• 1997 forecast: $13 bill

DAP Spr.‘98 ©UCB 40Interconnect TrendsNetwork>1000 m10 - 100 Mb/shigh (>ms)lowExtensive CRCChannel10 - 100 m40 - 1000 Mb/smediummediumByte Parit

DAP Spr.‘98 ©UCB 41Backplane Architectures128No16 - 32Single/MultipleMultipleNoAsync2512.927.913.621.5 mIEEE 101496Yes32Single/MultipleMultipleOptiona

DAP Spr.‘98 ©UCB 42Bus-Based Interconnect• Bus: a shared communication link between subsystems– Low cost: a single set of wires is shared multiple way

DAP Spr.‘98 ©UCB 43Bus-Based Interconnect• Two generic types of busses:– I/O busses: lengthy, many types of devices connected, wide range in the data

DAP Spr.‘98 ©UCB 44Bus Protocols° ° °Master SlaveControl LinesAddress LinesData LinesMultibus: 20 address, 16 data, 5 control, 50ns PauseBus Master:

DAP Spr.‘98 ©UCB 45Synchronous Bus Protocols AddressDataReadWaitClockAddressDataWaitPipelined/Split transaction Bus Protocoladdr 1data 0addr 2wait 1da

DAP Spr.‘98 ©UCB 46Asynchronous HandshakeAddressDataReadReq.Ack.Master Asserts AddressMaster Asserts DataNext AddressWrite Transactiont0 t1

DAP Spr.‘98 ©UCB 47Read TransactionAddressDataReadReqAckMaster Asserts Address Next Addresst0 t1 t2 t3 t4 t5Time Mul

DAP Spr.‘98 ©UCB 48Bus ArbitrationParallel (Centralized) ArbitrationSerial Arbitration (daisy chaining)PollingBR BGMBR BGMBR BGMMBGi BGoBRMBGi

DAP Spr.‘98 ©UCB 49Bus OptionsOption High performance Low costBus width Separate address Multiplex address& data lines & data linesData width

DAP Spr.‘98 ©UCB 5Review: Storage System Issues• Historical Context of Storage I/O• Secondary and Tertiary Storage Devices• Storage I/O Performance Me

DAP Spr.‘98 ©UCB 50SCSI: Small Computer System Interface• Clock rate: 5 MHz / 10 MHz (fast) / 20 MHz (ultra)• Width: n = 8 bits / 16 bits (wide); up

DAP Spr.‘98 ©UCB 51SCSI “Bus”: Channel ArchitectureCommand SetupArbitrationSelectionMessage Out (Identify)CommandDisconnect to seek/fill bufferMessage

DAP Spr.‘98 ©UCB 521993 I/O Bus Survey (P&H, 2nd Ed)Bus SBus TurboChannel MicroChannel PCIOriginator Sun DEC IBM IntelClock Rate (MHz) 16-25 12.5-

DAP Spr.‘98 ©UCB 531993 MP Server Memory Bus SurveyBus Summit Challenge XDBusOriginator HP SGI SunClock Rate (MHz) 60 48 66Split transaction? Yes Yes

DAP Spr.‘98 ©UCB 54Summary: I/O Benchmarks• Scaling to track technological change• TPC: price performance as nomalizing configuration feature• Auditin

DAP Spr.‘98 ©UCB 6ABCs of UNIX File Systems• Key Issues– File vs. Raw I/O– File Cache Size Policy– Write Policy– Local Disk vs. Server Disk• File vs.

DAP Spr.‘98 ©UCB 7ABCs of UNIX File Systems• Write Policy– File Storage should be permanent; either write immediately or flush file cache after fixed

DAP Spr.‘98 ©UCB 8ABCs of UNIX File Systems• Local vs. Server– Unix File systems have historically had different policies (and even file sytems) for l

DAP Spr.‘98 ©UCB 9Network File SystemsApplication ProgramUNIX System Call LayerUNIX File SystemBlock Device DriverVirtual File System InterfaceNFS Cli