System Organization amp Programming These slides are the product of many rounds of teaching CS 3410 by Deniz Altinbuke Kevin Walsh and Professors Weatherspoon Bala Bracy ID: 1030846
Download Presentation The PPT/PDF document "Storage CS 3410 Computer" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
1. StorageCS 3410Computer System Organization & ProgrammingThese slides are the product of many rounds of teaching CS 3410 by Deniz Altinbuke, Kevin Walsh, and Professors Weatherspoon, Bala, Bracy, and Sirer.
2. ChallengeHow do we store lots of data for a long timeDisk (Hard disk, floppy disk, …)Tape (cassettes, backup, VHS, …)CDs/DVDs
3. ChallengeHow do we store lots of data for a long timeDisk (Hard disk, floppy disk, …Solid State Disk (SSD)Tape (cassettes, backup, VHS, …)CDs/DVDsNon-Volitile Persistent Memory (NVM; e.g. 3D Xpoint)
4. I/O System CharacteristicsDependability is importantParticularly for storage devicesPerformance measuresLatency (response time)Throughput (bandwidth)Desktops & embedded systemsMainly interested in response time & diversity of devicesServersMainly interested in throughput & expandability of devices
5. Memory HierarchyTapeDiskDRAML2registers/L1100s, sequential access2 ns, random access5 ns, random access20-80 ns, random access2-8 ms, random access1 TB16 KB512 KB2 GB300 GB2006
6. Memory HierarchySSDDiskDRAML2registers/L1100ns-10us, random accessMillions of IOPS (I/O per sec)2 ns, random access5 ns, random access20-80 ns, random access2-8 ms, random access30 TB128 KB4 MB256 GB6 TB
7. Memory HierarchySSDDiskDRAML2registers/L1100ns-10us, random accessMillions of IOPS (I/O per sec)2 ns, random access5 ns, random access20-80 ns, random access2-8 ms, random access30 TB128 KB4 MB256 GB6 TBNon-volatile memory20 -100 ns, random access1 TB
8. Memory HierarchySSD100ns-10us, random accessMillions of IOPS (I/O per sec)30 TBServer10s of Disks256 TBRack of Servers10s of Servers10 PBData Center10-100s of Servers1 EBCloud10-100s of Data Centers0.1 YB245B248B253B260B267B
9. How big is Big Data in the Cloud?Exabytes: Delivery of petabytes of storage dailyTitan tech boom, randy katz, 2008The Rise of Cloud Computing
10. How big is Big Data in the Cloud?Most of the worlds data (and computation) hosted by few companies The Rise of Cloud Computing
11. How big is Big Data in the Cloud?Most of the worlds data (and computation) hosted by few companies The Rise of Cloud Computing
12. The promise of the Cloudubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction.NIST Cloud DefinitionThe Rise of Cloud Computing
13. The promise of the Cloudubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction. NIST Cloud DefinitionThe Rise of Cloud Computing
14. Tapes Same basic principle for 8-tracks, cassettes, VHS, ...Ferric Oxide Powder: ferromagnetic materialDuring recording, the audio signal is sent through the coil of wire to create a magnetic field in the core.During playback, the motion of the tape creates a varying magnetic field in the core and therefore a signal in the coil.0 0 1 0 1 0 1 0 1
15. Disks & CDs Disks use same magnetic medium as tapesconcentric rings (not a spiral)CDs & DVDs use optics and a single spiral track
16. Disk PhysicsTypical parameters :1 spindle1 arm assembly1-4 platters1-2 sides/platter1 head per side (but only 1 active head at a time) 700-20480 tracks/surface 16-1600 sectors/track
17. Disk AccessesAccessing a disk requires:specify sector: C (cylinder), H (head), and S (sector)specify size: number of sectors to read or writespecify memory addressPerformance:seek time: move the arm assembly to trackRotational delay: wait for sector to come aroundtransfer time: get the bits off the diskController time: time for setupTrackSectorSeek TimeRotationDelay
18. ExampleAverage time to read/write 512-byte sectorDisk rotation at 10,000 RPMSeek time: 6msTransfer rate: 50 MB/secController overhead: 0.2 msAverage time:Seek time + rotational delay + transfer time + controller overhead6ms + 0.5 rotation/(10,000 RPM) + 0.5KB/(50 MB/sec) + 0.2ms6.0 + 3.0 + 0.01 + 0.2 = 9.2ms
19. Disk Access ExampleIf actual average seek time is 2msAverage read time = 5.2ms
20. Disk SchedulingGoal: minimize seek timesecondary goal: minimize rotational latencyFCFS (First come first served)Shortest seek time SCAN/ElevatorFirst service all requests in one directionThen reverse and serve in opposite directionCircular SCANGo off the edge and come to the beginning and start all over again
21. FCFS
22. SSTF
23. SCAN
24. C-SCAN
25. Disk Geometry: LBA New machines use logical block addressing instead of CHSmachine presents illusion of an array of blocks, numbered 0 to N Modern disks…have varying number of sectors per trackroughly constant data density over diskvarying throughput over diskremap and reorder blocks (to avoid defects)completely obscure their actual physical geometryhave built-in caches to hide latencies when possible (but being careful of persistence requirements)have internal software running on an embedded CPU
26. Flash StorageNonvolatile semiconductor storage100× – 1000× faster than diskSmaller, lower powerBut more $/GB (between disk and DRAM)But, price is dropping and performance is increasing faster than disk
27. Flash TypesNOR flash: bit cell like a NOR gateRandom read/write accessUsed for instruction memory in embedded systemsNAND flash: bit cell like a NAND gateDenser (bits/area), but block-at-a-time accessCheaper per GBUsed for USB keys, media storage, …Flash bits wears out after 1000’s of accessesNot suitable for direct RAM or disk replacementFlash has unusual interfacecan only “reset” bits in large blocks
28. I/O vs. CPU PerformanceAmdahl’s LawDon’t neglect I/O performance as parallelism increases compute performanceExampleBenchmark takes 90s CPU time, 10s I/O timeDouble the number of CPUs/2 yearsI/O unchangedYearCPU timeI/O timeElapsed time% I/O timenow90s10s100s10%+245s10s55s18%+423s10s33s31%+611s10s21s47%
29. RAIDRedundant Arrays of Inexpensive DisksBig idea:Parallelism to gain performanceRedundancy to gain reliability
30. Raid 0StripingNon-redundant disk array!
31. Raid 1Mirrored Disks!More expensiveOn failure use the extra copy
32. Raid 2-3-4-5-6Bit Level Striping and Parity Checks!As level increases:More guarantee against failure, more reliabilityBetter read/write performanceRaid 2Raid 3Raid 4Raid 5
33. SummaryDisks provide nonvolatile memoryI/O performance measuresThroughput, response timeDependability and cost very importantRAIDRedundancy for fault tolerance and speed