Page 2 - Review: A Little Queuing Theory; Length
DAP Spr.‘98 ©UCB 2 Review: A Little Queuing Theory • Queuing models assume state of equilibrium: input rate = output rate • Notation: r average number of arriving customers/second T ser average time to service a customer (tradtionally µ = 1/ T ser ) u server utilization (0..1): u = r x T ser T q ave...
Page 3 - Review: Redundant Arrays of
DAP Spr.‘98 ©UCB 3 Review: Redundant Arrays of Disks (RAID) Techniques • Disk Mirroring, Shadowing (RAID 1) Each disk is fully duplicated onto its "shadow" Logical write = two physical writes 100% capacity overhead • Parity Data Bandwidth Array (RAID 3) Parity computed horizontally Logically...
Page 4 - RAID sales
DAP Spr.‘98 ©UCB 4 Review: RAID RAID sales • 1993: $3.4 billion on 214,667 arrays ( ≈ $15,000 / RAID) • 1996 forecast: $11 billion • 1997 forecast: $13 billion on 837,155 units – Source: DISK/TREND, 5/94 (415-961-6209)
Page 5 - Review: Storage System Issues; Historical Context of Storage I/O; A Little Queuing Theory
DAP Spr.‘98 ©UCB 5 Review: Storage System Issues • Historical Context of Storage I/O • Secondary and Tertiary Storage Devices • Storage I/O Performance Measures • Processor Interface Issues • A Little Queuing Theory • Redundant Arrarys of Inexpensive Disks (RAID) • ABCs of UNIX File Systems • I/O Be...
Page 6 - ABCs of UNIX File Systems; Key Issues; File; File Cache Size Policy
DAP Spr.‘98 ©UCB 6 ABCs of UNIX File Systems • Key Issues – File vs. Raw I/O – File Cache Size Policy – Write Policy – Local Disk vs. Server Disk • File vs. Raw: – File system access is the norm: standard policies apply – Raw: alternate I/O system to avoid file system, used by data bases • File Cach...
Page 7 - Write Policy
DAP Spr.‘98 ©UCB 7 ABCs of UNIX File Systems • Write Policy – File Storage should be permanent; either write immediately or flush file cache after fixed period (e.g., 30 seconds) – Write Through with Write Buffer – Write Back – Write Buffer often confused with Write Back » Write Through with Write B...
Page 9 - Network File Systems
DAP Spr.‘98 ©UCB 9 Network File Systems Application Program UNIX System Call Layer UNIX File System Block Device Driver Virtual File System Interface NFS Client Network Protocol Stack UNIX System Call Layer Virtual File System Interface NFS File System RPC/Transmission Protocols UNIX System Call Lay...
Page 10 - Typical File Server Architecture; Limits to performance: data copying; No specialization for fast processing between network
DAP Spr.‘98 ©UCB 10 Typical File Server Architecture Kernel NFS Protocol & File Processing TCP/IP Protocols Unix File System Ethernet Ethernet Driver Disk Manager & Driver Primary Memory Disk Controller Backplane Bus ... NFSRequest Single Processor File Server Limits to performance: data cop...
Page 11 - • Special hardware/software architecture for high
DAP Spr.‘98 ©UCB 11 AUSPEX NS5000 File Server Primary Memory Primary Memory Host Processor Host Memory Ethernet Processor File Processor Storage Processor Enhanced VME Backplane ... 1 2 10 Parallel SCSI Channels File Processor Ethernet Processor Independent File System Single Board Computer • Specia...
Page 12 - AUSPEX Software Architecture
DAP Spr.‘98 ©UCB 12 AUSPEX Software Architecture Unix System Call Layer VFS Interface NFS Client LFS Client Host Processor LFS Client NSF Server Protocols Network I/F Ethernet Processor LFS Server File System Server File Processor Primary Memory Storage Processor Disk Arrays Ethernet Primary control...
Page 13 - Berkeley RAID-II Disk Array File Server
DAP Spr.‘98 ©UCB 13 Berkeley RAID-II Disk Array File Server 8 Port Interleaved Memory (128 MByte) 8 x 8 x 32-bit Crossbar VME VME VME VME VME XOR X-BusBoard 8 Port Interleaved Memory (128 MByte) 8 x 8 x 32-bit Crossbar VME VME VME VME VME XOR IOPB In IOPB Out X-BusBoard HiPPIS HiPPID HiPPI TMCIOP Bu...
Page 14 - Email URL of initial project home page to TA?
DAP Spr.‘98 ©UCB 14 CS 252 Administrivia • Email URL of initial project home page to TA? • Upcoming events in CS 252 19-Mar Thu Send in Project Survey #2 20-Mar Fri Computer Pioneers and Pioneer Computers (Video, Gordon Bell Host) 23-Mar to 27-Mar Spring Break
Page 15 - For better or worse, benchmarks shape a field
DAP Spr.‘98 ©UCB 15 I/O Benchmarks • For better or worse, benchmarks shape a field – Processor benchmarks classically aimed at response time for fixed sized problem – I/O benchmarks typically measure throughput, possibly with upper limit on response times (or 90% of response times) • What if fix pro...
Page 16 - self-scaling benchmark; Describe three self-scaling benchmarks
DAP Spr.‘98 ©UCB 16 I/O Benchmarks • Alternative: self-scaling benchmark ; automatically and dynamically increase aspects of workload to match characteristics of system measured – Measures wide range of current & future • Describe three self-scaling benchmarks – Transacition Processing: TPC-A, T...
Page 17 - Processing; Atomic
DAP Spr.‘98 ©UCB 17 I/O Benchmarks: Transaction Processing • Transaction Processing (TP) (or On-line TP=OLTP) – Changes to a large body of shared information from many terminals, with the TP system guaranteeing proper behavior on a failure – If a bank’s computer fails when a customer withdraws money...
Page 18 - Early 1980s great interest in OLTP
DAP Spr.‘98 ©UCB 18 I/O Benchmarks: Transaction Processing • Early 1980s great interest in OLTP – Expecting demand for high TPS (e.g., ATM machines, credit cards) – Tandem’s success implied medium range OLTP expands – Each vendor picked own conditions for TPS claims, report only CPU times with widel...
Page 20 - second
DAP Spr.‘98 ©UCB 20 I/O Benchmarks: TP1 by Anon et. al • DebitCredit Scalability: size of account, branch, teller, history function of throughput TPS Number of ATMs Account-file size 10 1,000 0.1 GB 100 10,000 1.0 GB 1,000 100,000 10.0 GB 10,000 1,000,000 100.0 GB – Each input TPS =>100,000 accou...
Page 21 - Problems; Solution: Hire auditor to certify results
DAP Spr.‘98 ©UCB 21 I/O Benchmarks: TP1 by Anon et. al • Problems – Often ignored the user network to terminals – Used transaction generator with no think time; made sense for database vendors, but not what customer would see • Solution: Hire auditor to certify results – Auditors soon saw many varia...
Page 22 - These have been withdrawn as benchmarks
DAP Spr.‘98 ©UCB 22 I/O Benchmarks: Old TPC Benchmarks • TPC-A: Revised version of TP1/DebitCredit – Arrivals: Random (TPC) vs. uniform (TP1) – Terminals: Smart vs. dumb (affects instruction path length) – ATM scaling: 10 terminals per TPS vs. 100 – Branch scaling: 1 branch record per TPS vs. 10 – R...
Page 23 - Models a wholesale supplier managing orders
DAP Spr.‘98 ©UCB 23 I/O Benchmarks: TPC-C Complex OLTP • Models a wholesale supplier managing orders • Order-entry conceptual model for benchmark • Workload = 5 transaction types • Users and database scale linearly with throughput • Defines full-screen end-user interface • Metrics: new-order rate (t...
Page 24 - Decision Support Workload; OLTP: business operation
DAP Spr.‘98 ©UCB 24 I/O Benchmarks: TPC-D Complex Decision Support Workload • OLTP: business operation • Decision support: business analysis (historical) • Workload = 17 adhoc transactions – e,g., Impact on revenue of eliminating company-wide discount? • Synthetic generator of data • Size determined...
Page 25 - Transactional Web Benchmark
DAP Spr.‘98 ©UCB 25 I/O Benchmarks: TPC-W Transactional Web Benchmark • Represent any business (retail store, software distribution, airline reservation, electronic stock trades, etc.) that markets and sells over the Internet/ Intranet • Measure systems supporting users browsing, ordering, and condu...
Page 29 - TPC-D Performance 1TB
DAP Spr.‘98 ©UCB 29 TPC-D Performance 1TB Rank Config. Qppd QthD $/QphD Database 1 Sun Ultra E6000 (4 x 24-way) 12,931.9 5,850.3 1,353.00 Infomix Dyn 2 NCR WorldMark (32 x 4-way) 12,149.2 3,912.3 2103.00 Teradata 3 IBM RS/6000 SP (32 x 8-way) 7,633.0 5,155.4 2095.00 DB2 UDB, V5 • NOTE: Inappropriate...
Page 31 - Predecessor: NFSstones
DAP Spr.‘98 ©UCB 31 SPEC SFS/LADDIS Predecessor: NFSstones • NFSStones: synthetic benchmark that generates series of NFS requests from single client to test server: reads, writes, & commands & file sizes from other studies – Problem: 1 client could not always stress server – Files and block ...
Page 33 - Example SPEC SFS Result: DEC Alpha; FDDI networks 32 NFS Daemons, 24 GB file size
DAP Spr.‘98 ©UCB 33 Example 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 Daemons, 24 GB file size • 88 Disks, 16 controllers, 84 file systems NFS Throughput (nfs ops/sec) Avg. NSF Resp. Time 0 10 20 30 40 50 0 1000 200...
Page 34 - Willy; Self scaling to automatically explore system size; Fix four parameters while vary one parameter
DAP Spr.‘98 ©UCB 34 Willy • UNIX File System Benchmark that gives insight into I/O system behavior (Chen and Patterson, 1993) • Self scaling to automatically explore system size • Examines five parameters – Unique bytes touched : ≈ data size; locality via LRU » Gives file cache size – Percentage of ...
Page 35 - Sprite
DAP Spr.‘98 ©UCB 35 Example Willy: DS 5000 Sprite Ultrix Avg. Access Size 32 KB 13 KB Data touched (file cache) 2MB, 15 MB 2 MB Data touched (disk) 36 MB 6 MB • % reads = 50%, % sequential = 50% • DS 5000 32 MB memory • Ultrix: Fixed File Cache Size, Write through • Sprite: Dynamic File Cache Size, ...
Page 36 - Sprite's Log Structured; Write-Optimized File System
DAP Spr.‘98 ©UCB 36 Sprite's Log Structured File System Large file caches effective in reducing disk reads Disk traffic likely to be dominated by writes Write-Optimized File System • Only representation on disk is log • Stream out files, directories, maps without seeks Advantages: • Speed• Stripes e...
Page 37 - Willy: DS 5000 Number Bytes Touched
DAP Spr.‘98 ©UCB 37 Willy: DS 5000 Number Bytes Touched • Log Structured File System: effective write cache of LFS much smaller (5-8 MB) than read cache (20 MB) – Reads cached while writes are not => 3 plateaus Number MB Touched MB/sec 0 1 2 3 4 5 6 7 8 1 10 100 Sprite U l t r i x
Page 38 - Scaling to track technological change
DAP Spr.‘98 ©UCB 38 Summary: I/O Benchmarks • Scaling to track technological change • TPC: price performance as nomalizing configuration feature • Auditing to ensure no foul play • Througput with restricted response time is normal measure
Page 40 - Interconnect Trends; High speed hardware interfaces + logical protocols
DAP Spr.‘98 ©UCB 40 Interconnect Trends Network >1000 m 10 - 100 Mb/s high (>ms) low Extensive CRC Channel 10 - 100 m 40 - 1000 Mb/s medium medium Byte Parity Backplane 1 m 320 - 1000+ Mb/s low (<µs) high Byte Parity Distance Bandwidth Latency Reliability • Interconnect = glue that interfac...
Page 41 - Backplane Architectures; Distinctions begin to blur:
DAP Spr.‘98 ©UCB 41 Backplane Architectures 128 No 16 - 32 Single/Multiple Multiple No Async 25 12.927.913.6 21 .5 m IEEE 1014 96 Yes 32 Single/Multiple Multiple Optional Async 37 15.595.220.8 20 .5 m IEEE 896 Metric VME FutureBus 96 Yes 32 Single/Multiple Multiple Optional Sync 201040 13.3 21 .5 m ...
Page 42 - Bus-Based Interconnect; Bus: a shared communication link between subsystems; Disadvantage
DAP Spr.‘98 ©UCB 42 Bus-Based Interconnect • Bus: a shared communication link between subsystems – Low cost: a single set of wires is shared multiple ways – Versatility: Easy to add new devices & peripherals may even be ported between computers using common bus • Disadvantage – A communication b...
Page 43 - Two generic types of busses:; Bus transaction
DAP Spr.‘98 ©UCB 43 Bus-Based Interconnect • Two generic types of busses: – I/O busses: lengthy, many types of devices connected, wide range in the data bandwidth), and follow a bus standard(sometimes called a channel ) – CPU–memory buses: high speed, matched to the memory system to maximize memory–...
Page 44 - Bus Protocols; Bus Master
DAP Spr.‘98 ©UCB 44 Bus Protocols ° ° ° Master Slave Control LinesAddress LinesData Lines Multibus: 20 address, 16 data, 5 control, 50ns Pause Bus Master : has ability to control the bus, initiates transaction Bus Slave : module activated by the transaction Bus Communication Protocol : specification...
Page 45 - Synchronous Bus Protocols
DAP Spr.‘98 ©UCB 45 Synchronous Bus Protocols Address Data Read Wait Clock Address Data Wait Pipelined/Split transaction Bus Protocol addr 1 data 0 addr 2 wait 1 data 1 addr 3 OK 1 data 2 begin read Read complete
Page 46 - Asynchronous Handshake; Write Transaction
DAP Spr.‘98 ©UCB 46 Asynchronous Handshake Address Data Read Req. Ack. Master Asserts Address Master Asserts Data Next Address Write Transaction t0 t1 t2 t3 t4 t5 t0 : Master has obtained control and asserts address, direction, data Waits a specified amount of time for slaves to decode target\ t1: M...
Page 47 - Read Transaction; Time Multiplexed Bus: address and data share lines
DAP Spr.‘98 ©UCB 47 Read Transaction Address Data Read Req Ack Master Asserts Address Next Address t0 t1 t2 t3 t4 t5 Time Multiplexed Bus: address and data share lines t0 : Master has obtained control and asserts address, direction, data Waits a specified amount of time for slaves to decode target\ ...
Page 48 - Bus Arbitration
DAP Spr.‘98 ©UCB 48 Bus Arbitration Parallel (Centralized) Arbitration Serial Arbitration (daisy chaining) Polling BR BG M BR BG M BR BG M M BGi BGo BR M BGi BGo BR M BGi BGo BR BG BR A.U. BR A C M BR A C M BR A C M BRA A.U. Bus RequestBus Grant
Page 49 - Bus Options
DAP Spr.‘98 ©UCB 49 Bus Options Option High performance Low cost Bus width Separate address Multiplex address & data lines & data lines Data width Wider is faster Narrower is cheaper (e.g., 32 bits) (e.g., 8 bits) Transfer size Multiple words has Single-word transfer less bus overhead is sim...
Page 50 - SCSI: Small Computer System Interface
DAP Spr.‘98 ©UCB 50 SCSI: Small Computer System Interface • Clock rate: 5 MHz / 10 MHz (fast) / 20 MHz (ultra) • Width: n = 8 bits / 16 bits (wide); up to n – 1 devices to communicate on a bus or “string” • Devices can be slave (“target”) or master(“initiator”) • SCSI protocol: a series of “phases”,...
Page 51 - SCSI “Bus”: Channel Architecture
DAP Spr.‘98 ©UCB 51 SCSI “Bus”: Channel Architecture Command Setup Arbitration Selection Message Out (Identify) Command Disconnect to seek/fill buffer Message In (Disconnect) - - Bus Free - - Arbitration Reselection Message In (Identify) Data Transfer Data In Disconnect to fill buffer Message In (Sa...
Page 53 - 993 MP Server Memory Bus Survey
DAP Spr.‘98 ©UCB 53 1993 MP Server Memory Bus Survey Bus Summit Challenge XDBus Originator HP SGI Sun Clock Rate (MHz) 60 48 66 Split transaction? Yes Yes Yes? Address lines 48 40 ?? Data lines 128 256 144 (parity) Data Sizes (bits) 512 1024 512 Clocks/transfer 4 5 4? Peak (MB/s) 960 1200 1056 Maste...
