Introduction to I/O

1. ENGS 116 Lecture 17 Vincent Berk November 19, 2008 Reading for today: Sections 4.4 – 4.9 Reading for Monday: Sections 6.1 – 6.4 Reading for next Monday: Sections 6.5 – 6.9 Homework for Friday: 5.4, 5.6, 5.10, 4.1, 4.17 Introduction to I/O

2. ENGS 116 Lecture 17 Processor Control Input Memory Datapath Output The Big Picture: Where are we now?

3. ENGS 116 Lecture 17 I/O Systems

4. ENGS 116 Lecture 17 Storage System Issues • Historical Context of Storage I/O • Secondary and Tertiary Storage Devices • Storage I/O Performance Measures • A Little Queueing Theory • Processor Interface Issues • I/O Buses • Redundant Arrays of Inexpensive Disks (RAID) • I/O Benchmarks

5. ENGS 116 Lecture 17 ENGS 116 Lecture 17 5 Motivation: Who Cares About I/O? • CPU Performance: 50% to 100% per year • I/O system performance limited by mechanical delays < 5% per year (IO per sec or MB per sec) • Amdahl's Law: system speed-up limited by the slowest part! 10% IO & 10x CPU => 5x Performance (lose 50%) 10% IO & 100x CPU => 10x Performance (lose 90%) • I/O bottleneck: Diminishing fraction of time in CPU Diminishing value of faster CPUs

6. ENGS 116 Lecture 17 The I/O GAP ENGS 116 Lecture 17 6 Technology Trends • Today: Processing power doubles every 18 months •  Today: Memory size doubles every 18 months (4X/3 yrs) •  Today: Disk capacity doubles every 18 months • Disk positioning rate (seek + rotate) doubles every ten years!

7. ENGS 116 Lecture 17 Storage Technology Drivers • Driven by the prevailing computing paradigm • 1950s: migration from batch to on-line processing • 1990s: migration to ubiquitous computing • computers in phones, books, cars, video cameras, … • nationwide fiber optical network with wireless tails • Effects on storage industry: • Embedded storage • smaller, cheaper, more reliable, lower power • Interesting twist: Apple I-pod • Data utilities • high capacity, hierarchically managed storage

8. ENGS 116 Lecture 17 Historical Perspective • 1956 IBM Ramac — early 1970s Winchester • Developed for mainframe computers, proprietary interfaces • Steady shrink in form factor: 27 in. to 14 in. • 1970s developments • 5.25-inch floppy disk form factor • early emergence of industry standard disk interfaces • ST506, SASI, SMD, ESDI • Early 1980s • PCs and first generation workstations • Mid 1980s • Client/server computing • Centralized storage on file server • accelerates disk downsizing: 8 inch to 5.25 inch • Mass market disk drives become a reality • industry standards: SCSI, IDE, SATA • 5.25-inch drives for standalone PCs, end of proprietary interfaces

9. ENGS 116 Lecture 17 Disk History Data density Mbit/sq. in. Capacity of Unit Shown Megabytes 1973: 1. 7 Mbit/sq. in 140 MBytes 1979: 7. 7 Mbit/sq. in 2,300 MBytes Source: New York Times, 2/23/98, page C3, “Makers of disk drives crowd even more data into even smaller spaces”

10. ENGS 116 Lecture 17 Historical Perspective • Late 1980s/Early 1990s: • Laptops, notebooks, (palmtops) • 3.5 inch, 2.5 inch, 1.8 inch formfactors • Formfactor plus capacity drives market, not so much performance • Recently Bandwidth improving at 40%/ year • Challenged by DRAM, flash RAM in PCMCIA cards • still expensive, Intel promises but doesn’t deliver • unattractive MBytes per cubic inch • Optical disk fails on performance (e.g., NEXT) but finds niche (CD ROM)

11. ENGS 116 Lecture 17 Disk History 1989: 63 Mbit/sq. in 60,000 MBytes 1997: 1450 Mbit/sq. in 2300 MBytes 1997: 3090 Mbit/sq. in 8100 MBytes Source: New York Times, 2/23/98, page C3, “Makers of disk drives crowd even more data into even smaller spaces”

12. ENGS 116 Lecture 17 ENGS 116 Lecture 17 12 Alternative Data Storage Technologies: Early 1990s Data Cap BPI TPI BPI*TPI Transfer Access Technology (MB) (Million) (KByte/s) Time Conventional Tape: Cartridge (.25") 150 12000 104 1.2 92 minutes IBM 3490 (.5") 800 22860 38 0.9 3000 seconds Helical Scan Tape: Video (8mm) 4600 43200 1638 71 492 45 secs DAT (4mm) 1300 61000 1870 114 183 20 secs D-3 (1/2") 20,000 15 secs? Magnetic & Optical Disk: Hard Disk (5.25") 1200 33528 1880 63 3000 18 ms IBM 3390 (10.5") 3800 27940 2235 62 4250 20 ms Sony MO (5.25") 640 24130 18796 454 88 100 ms

13. ENGS 116 Lecture 17 Track Sector Cylinder Platter Head Response time = Queue + Controller + Seek + Rot + Transfer Service time ENGS 116 Lecture 17 13 Devices: Magnetic Disks • Purpose: • Long-term, nonvolatile storage • Large, inexpensive, slow level in the storage hierarchy • Characteristics: • Seek Time (~ 8 ms avg) • positional latency • rotational latency • Transfer rate • About a sector per ms (10-40 MB/s) • Blocks • Capacity • Gigabytes • Quadruples every 3 years 7200 RPM = 120 RPS  8 ms per rev avg. rot. latency = 4 ms 128 sectors per track  0.0625 ms per sector 1 KB per sector  16 MB / s

14. ENGS 116 Lecture 17 Disk Latency = Queueing Time + Controller Time + Seek Time + Rotation Time + Transfer Time Order-of-magnitude times for 4K byte transfers: Seek: 8 ms or less Rotate: 4.2 ms @ 7200 rpm Transfer: 1 ms @ 7200 rpm ENGS 116 Lecture 17 14 Disk Device Terminology Sector Head Inner Track Outer Track Platter Arm Actuator

15. ENGS 116 Lecture 17 ENGS 116 Lecture 17 15 Disk Device Terminology

16. ENGS 116 Lecture 17 Tape vs. Disk • Longitudinal tape uses same technology as hard disk; tracks its density improvements • Disk head flies above surface, tape head lies on surface • Inherent cost-performance based on geometries: fixed rotating platters with gaps (random access, limited area, 1 media / reader) vs. removable long strips wound on spool (sequential access, "unlimited" length, multiple / reader) • Modern technology: Helical Scan (VCR, Camcorder, DAT) Spins head at angle to tape to improve density

17. ENGS 116 Lecture 17 R-DAT Technology Rotary Drum R W 2000 RPM W R 90° Wrap Angle Direction Drum of Tape Track Four Head Recording Tracks Recorded ± 20° w/o guard band Read After Write Verify Helical Recording Scheme

18. ENGS 116 Lecture 17 Example: R-DAT Technology • Rotating (vs. Stationary) head (Digital Audio Tape) • Highest areal recording density commercially available • High density due to: • high coercivity metal tape • helical scan recording method • narrow, gapless (overlapping) recording tracks • 10X improvement capacity & transfer rate by 1999 • faster tape and drum speeds • greater track overlap • No significant changes since late 90s

19. ENGS 116 Lecture 17 DDS ANSI Standard (HP, SONY) Track Tape Frame R-DAT Technology 65% of Track is Data Area 70% Data Bytes 30% Bytes Parity Plus Reed-Solomon Codes Track Finding Area (Servo) Subcode Area (Index) Margin Area Block Track (2900 Data Bytes) Frame (2 Tracks) Group (22 Frames + Optional Group ECC, 128 K bytes) Theoretical Bit Error Rates: • w/o group ECC: one in 1026 • w/ group ECC: one in 1033

20. ENGS 116 Lecture 17 ENGS 116 Lecture 17 20 Current Drawbacks to Tape • Tape wears out: • Helical 100s of passes to 1000s for longitudinal • Head wears out: • 2000 hours for helical • Both must be accounted for in economic/reliability model • Long rewind, eject, load, spin-up times; not inherent, just no need in marketplace (so far) • Designed for archival storage • Capacities: 700 GB per tape, but used in big libraries (150 PB)

21. ENGS 116 Lecture 17 Modern Day Tape specs. • Sun StorageTek SL 8500 tape library • Up to 150,000 TB of archival storage • With 2,048 drives: 884.7 TB/hr • 60 minutes to read entire library • 11 seconds to find and load any type in library

22. ENGS 116 Lecture 17 Optical Disks: CD-ROM vs. DVD • Both 12 cm in diameter (most-common size for CD-ROM) • CD-ROM • Maximum 80 minutes of music • 700 MB of information • DVD (Digital Versatile/Video Disk) • 4.7 GB of information on one of its two sides — enough for 133-minute movie • With 2 layers on each of its two sides, will hold up to 17 gigabytes • Uses MPEG-2 file compression standard • Blue Ray DVD (higher density) • Up to 200 GB of capacity per disc

23. ENGS 116 Lecture 17 Queue Proc IOC Device Response time = Queue + Device Service time Disk I/OPerformance Metrics: Response Time Throughput

24. ENGS 116 Lecture 17 Response Time vs. Productivity • Interactive environments: Each interaction or transactionhas 3 parts: • Entry Time: time for user to enter command • System Response Time: time between user entry & system replies • Think Time: Time from response until user begins next command 1st transaction 2nd transaction • What happens to transaction time as system response time shrinks from 1.0 sec to 0.3 sec? • With keyboard: 4.0 sec entry, 9.4 sec think time • With graphics: 0.25 sec entry, 1.6 sec think time

25. ENGS 116 Lecture 17 Response Time & Productivity conventional 0.3s • 0.7 sec off response saves 4.9 sec (34%) and 2.0 sec (70%) total time per transaction => greater productivity • Another study: everyone gets more done with faster response, but novice with fast response = expert with slow conventional 1.0s graphics 0.3s entry resp think graphics 1.0s 0.00 5.00 10.00 15.00 Time

26. ENGS 116 Lecture 17 Response Time & Productivity

27. ENGS 116 Lecture 17 Disk Time Example • Disk parameters: • Transfer size is 8K bytes • Advertised average seek is 12 ms • Disk spins at 7200 RPM • Transfer rate is 4 MB/sec • Controller overhead is 2 ms • Assume that disk is idle so no queueing delay • What is average disk access time for a sector? • Avg seek + avg rotational delay + transfer time + controller overhead • 12 ms + 0.5/(7200 RPM/60) + 8 KB/4 MB/s + 2 ms • 12 + 4.15 + 2 + 2 = 20 ms • Advertised seek time assumes no locality: typically 1/4 to 1/3 advertised seek time: 20 ms => 12 ms

28. ENGS 116 Lecture 17 The following simplified diagram shows two potential ways of numbering the sectors of data on a disk (only two tracks are shown and each track has eight sectors). Assuming that typical reads are contiguous (e.g., all 16 sectors are read in order), which way of numbering the sectors will be likely to result in higher performance? Why? 0 0 1 1 7 7 8 14 9 15 15 13 2 2 10 8 14 12 6 6 9 11 11 13 10 12 3 3 5 5 4 4

29. ENGS 116 Lecture 17 Processor Interface Issues • Interconnections • Busses • Processor Interface • Interrupts • Memory mapped I/O • I/O Control Structures • Polling • Interrupts • DMA • I/O controllers • I/O processors • Capacity, Access Time, Bandwidth

30. ENGS 116 Lecture 17 Busses • Bus: a shared communication link between subsystems • Disadvantage: a communication bottleneck, possibly limiting the maximum I/O throughput • Bus speed is limited by physical factors • Two generic types of busses • I/O • CPU-memory • Bus transaction: sending address & receiving or sending data

31. ENGS 116 Lecture 17 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 simpler Bus masters Multiple Single master (requires arbitration) (no arbitration) Split Yes — separate No — continuous transaction? Request and Reply connection is cheaper packets for higher and has lower latency bandwidth (needs multiple masters) Clocking Synchronous Asynchronous

32. ENGS 116 Lecture 17 CPU Memory Independent I/O bus Memory bus Interface Interface Peripheral Peripheral CPU Memory Interface Interface Peripheral Peripheral I/O Interface Separate I/O instructions (in, out) Common memory & I/O bus 40 Mbytes/sec optimistically VME bus Multibus-II Nubus 10 MIP processor completely saturates the bus!

33. ENGS 116 Lecture 17 Backplane bus Memory Processor a. I/O devices Processor-memory bus Memory Processor Bus adapter Bus adapter Bus adapter I/O bus I/O bus I/O bus b. Processor-memory bus Memory Processor Bus adapter I/O bus Bus adapter Backplane bus Bus adapter I/O bus c.