Presentation is loading. Please wait.

Presentation is loading. Please wait.

Prof. John Nestor ECE Department Lafayette College Easton, Pennsylvania 18042 ECE 313 - Computer Organization Lecture 22 - Input/Output.

Similar presentations


Presentation on theme: "Prof. John Nestor ECE Department Lafayette College Easton, Pennsylvania 18042 ECE 313 - Computer Organization Lecture 22 - Input/Output."— Presentation transcript:

1 Prof. John Nestor ECE Department Lafayette College Easton, Pennsylvania 18042 nestorj@lafayette.edu ECE 313 - Computer Organization Lecture 22 - Input/Output Fall 2004 Reading: Ch. 8 (Skim) Portions of these slides are derived from: Textbook figures © 1998 Morgan Kaufmann Publishers all rights reserved Tod Amon's COD2e Slides © 1998 Morgan Kaufmann Publishers all rights reserved Dave Patterson’s CS 152 Slides - Fall 1997 © UCB Rob Rutenbar’s 18-347 Slides - Fall 1999 CMU other sources as noted

2 ECE 313 Fall 2004Lecture 22 - Input / Output2 Roadmap for the term: major topics  Overview / Abstractions and Technology  Instruction sets  Logic & arithmetic  Performance  Processor Implementation  Single-cycle implemenatation  Multicycle implementation  Pipelined Implementation  Memory systems  Input/Output 

3 ECE 313 Fall 2004Lecture 22 - Input / Output3 Outline - Input-Output  Overview  I/O Performance Metrics  Types and Characteristics of I/O Devices  Buses  Interfacing I/O Devices

4 ECE 313 Fall 2004Lecture 22 - Input / Output4 Overview - Input Output  I/O is a critical part of computer systems  I/O has greater impact as processor performance increases (remember Amdahl’s law?) Figure 8.1 Cache Memory - I/O Bus Processor Main Memory I/O Controller I/O Controller I/O Controller Disk Graphics Network Interrupts

5 ECE 313 Fall 2004Lecture 22 - Input / Output5 Measuring I/O Performance  Individual transactions  Response time (Latency)  Bandwidth (Throughput) How much data can move through system in given time? … OR How many I/O operations can we do per unit time?  Overall Performance - Benchmarks for  Supercomputer I/O  Transaction Processing  File System I/O  Appropriate measure depends on application

6 ECE 313 Fall 2004Lecture 22 - Input / Output6 Characterizing I/O Devices  I/O Device Characteristics  Behavior: input only, output only, or storage  Partner: who or what is using the interface? Machine? Human Being?  Data Rate: peak rate of data transfer  Typical I/O Devices: See Fig. 8.2, p. 568, old p. 644

7 ECE 313 Fall 2004Lecture 22 - Input / Output7 Hard Disk Components Image Source: Seagate Technolgy LLC www.seagate.com Platters Spindle Arm Actuator Interface Connector (SCSI or EIDE) Electronics Head

8 ECE 313 Fall 2004Lecture 22 - Input / Output8 Hard Disk Organization Platters (2 sides each) Track Sector  Typical numbers (depending on the disk size):  500 to 2,000 tracks per surface  32 to 128 sectors per track A sector is the smallest unit that can be read or written  Traditionally all tracks have the same number of sectors:  Outer tracks record at lower bit density than inner tracks  Recent change: constant bit density, data rate varies with track

9 ECE 313 Fall 2004Lecture 22 - Input / Output9 Hard Disk Performance Factors  Seek Time  move head to desired track  today’s drives: 5-15ms  Rotational Latency  today’s drives: 5400-12,000RPM  average: 0.5 * ( 1 / RPM )  Transfer Time  time to transfer a sector (1KB)  today’s drives: 10-30MB/Sec  Controller Time  Overhead from on-drive electronics  added value: caching, prefetching Sector Track Head Platter Cylinder

10 ECE 313 Fall 2004Lecture 22 - Input / Output10 Hard Disk Examples Characteristics IBM 3090 IBM UltraStarIntegral 1820 Disk diameter (inches) 10.88 3.50 1.80 Formatted data capacity (MB)22,700 4,300 21 MTTF (hours)50,000 1,000,000 100,000 Number of arms/box 12 1 1 Rotation speed (RPM) 3,600 7,200 3,800 Transfer rate (MB/sec) 4.2 9-12 1.9 Power/box (watts) 2,900 13 2 MB/watt 8 102 10.5 Volume (cubic feet) 97 0.13 0.02 MB/cubic feet 234 33000 1050 See also book Fig. 8.4

11 ECE 313 Fall 2004Lecture 22 - Input / Output11 Buses  Goal: connect I/O Devices to Processor, Memory  Use shared wires for economic reasons  Challenging design problem due to physical constraints  length of bus  number of devices  Buses are specialized depending on purpose  Processor-Memoryshort / high-speed  Backplanemed. length / med. speed  I/Olong / low speed

12 ECE 313 Fall 2004Lecture 22 - Input / Output12 Buses Cache Processor Main Memory I/O Controller I/O Controller I/O Controller Disk Graphics Network Control Data  Bus Structure  Bus Operation  Control signals - used to start a transaction (input/output)  Data signals - used to transmit data (including addresses)

13 ECE 313 Fall 2004Lecture 22 - Input / Output13 Bus Variations  Synchronous vs. Asynchronous  Single master vs. multiple masters  Narrow vs. wide  Long vs. short

14 ECE 313 Fall 2004Lecture 22 - Input / Output14 Example Buses  Processor / Memory  Proprietary microprocessor buses - Max bandwidth 422 MByte/sec  Backplane  PCI - Max bandwidth 133 MByte/sec  ISA (obsolete, but still there in many systems)  I/O  SCSI - Max bandwidth 40 MByte/sec (60MByte/sec for 2.0)  USB - Max bandwidth 12 MByte/sec  FireWire (IEEE 1394) - Max bandwidth 50 Mbyte/sec (100MByte/sec for Firewire 800)

15 ECE 313 Fall 2004Lecture 22 - Input / Output15 Example - Buses on a Modern PC See also book Fig. 8.11-8.12

16 ECE 313 Fall 2004Lecture 22 - Input / Output16 Interfacing I/O - Processor-Device Communication  Processor accesses registers in device  Control Register  Status Register  Data Register  Connection styles  Memory-Mapped I/O registers mapped to memory locations accessed using load/store  I/O Instructions registers mapped to “device numbers” accessed using I/O instructions Processor Device Data Status Control

17 ECE 313 Fall 2004Lecture 22 - Input / Output17 Interfacing I/O - Processor-Device Communcation  Communicating w/ Processor  Polling processor checks device status register problem: wastes processor time  Interrupt-Driven I/O - initiated by device device interrupts processor processor interrupt handler processes input often combined with buffering  Direct Memory Access (DMA) Device takes over bus for block transfer Device interrupts processor when complete Cache issues: what if we transfer memory that’s stored in cache?

18 ECE 313 Fall 2004Lecture 22 - Input / Output18 I/O Summary  I/O has growing impact on performance  Wide range of I/O devices  Buses provide pathways for I/O connection  Software ultimately has to deal with I/O, too


Download ppt "Prof. John Nestor ECE Department Lafayette College Easton, Pennsylvania 18042 ECE 313 - Computer Organization Lecture 22 - Input/Output."

Similar presentations


Ads by Google