IO Management and Disk Scheduling

1
IO Management and Disk Scheduling

• B. Ramamurthy

2
Introduction

• In addition to primary memory (volatile),
  computer systems provide users secondary storage
  units which may be used for persistent (or
  permanent) storage.
• A file is a collection of data elements grouped
  together for the purposes of access control,
  retrieval and modification.
• A file system together with IO management
  subsystem accomplish the mapping of abstract user
  interface into the actual collection of hardware
  such as disk, tape etc. of varying
  characteristics.
• Recently there has been lot of interest in a
  storage medium called RAID. We will look into
  this (11.6)
• We will study IO management in this discussion
  (Ch.11) and File system in the next chapter (Ch.
  12)

3
Topics for discussion

• IO devices
• Device Characteristics
• IO Design Objectives
• A model of IO Organization
• Buffering
• Disk IO
• Disk scheduling
• Disk cache
• RAID storage
• Summary

4
IO Devices

• IO devices can be grouped into three categories
• Human readable Suitable for use with computer
  user. Example video display terminals, keyboard,
  mouse and printers.
• Machines readable Example disks, tapes,
  sensors, controllers.
• Communication Suitable for communicating with
  remote devices Drivers, modems, sockets, etc.

5
Device Characteristics

• Devices vary widely in characteristics such as
  Data rate, applications, complexity of control,
  unit of transfer, data representation, Error
  conditions.
• Programmed IO , Interrupt-driven IO and
  Direct-memory access are three techniques for
  performing IO.

6
IO Design Objectives

• Efficiency IO operations often form the
  bottleneck in a computing system.
• Generality Handle all devices in a uniform
  manner. How?
• Use hierarchical, modular approach to design of
  IO subsystem.
• This approach hides most of the details of device
  details in lower-levels so that processes and
  upper levels of OS see devices in terms of
  general functions such as Read, Write, Open,
  Lock, UnLock etc.

7
A Model of IO Organization
User Processes
User Processes
User Processes
Dir. Mgt.
Comm. Arch.
Logical IO
File System
Physical Org.
Device IO
Sched. Control
Hardware
Local Peripheral Device
Comm Port
File System
8
Buffering

• In order to avoid inefficiencies associated with
  direct transfer from user-process to IO device,
  buffering schemes can be used.
• Buffering smoothes out the peaks in IO demand.
• Buffering depends on the type of IO device
  Block-oriented (Ex disks) and Stream-oriented
  (terminal, comm. ports).
• Stream-oriented IO single buffering
• Block-oriented Devices double buffer, circular
  buffer

9
Disk IO

• A disk contains a stack of recording surfaces
  each surface with a movable read/write head.
• Each surface has circular tracks and each track
  divided into sectors. A stack of tracks make up a
  cylinder.
• To read or write, the head must be positioned at
  the desired track and at the beginning of the
  desired sector.

10
Disk IO

• Seek Time (Ts) is the time it takes to position
  the head at the desired track.
• Rotational Delay (Tr) is the time it takes to to
  line up the head at the beginning of the desired
  sector.
• Access Time (Ta ) Ts Tr
• The data transfer time (Tt) is the time to
  read/write data at the desired position.

11
Disk Scheduling

• Among the times Seek time is the most critical.
  Average access time can be improved by reducing
  the average Seek Time.
• Assumption A queue of requests for each IO
  device is maintained. At any time a number of
  requests from various processes are in the queue.
  Consider a sequence of n-requests for scheduling.
• Example Tracks 55, 58, 39, 18, 90, 160, 150, 38
• Using FIFO, SSTF, SCAN, C-SCAN. Table 11.3,
  Fig.11.7

12
Disk Cache

• A disk cache is a buffer in main memory for disk
  sectors.
• Cache contains a copy of the some of the sectors
  on the disk.
• When a IO request is made for a particular
  sector, if the data is available on disk cache it
  is taken. Otherwise it can be brought into the
  cache from the disk.
• Similar to virtual memory principles. Directory
  blocks are good candidates for caching.

13
Redundant Array of Independent Disks (RAID)

• Is a set of disk drives viewed by operating
  system as a single logical device.
• Data are distributed across the physical drives
  of an array.
• Redundant disk capacity is used to store parity
  information, which guarantees data recoverability
  in case of disk failure.
• Raid levels 0 to 7

14
RAID Levels

• RAID 0 - striping
• RAID 1 - Mirrored
• RAID 2 - Error correcting using Hamming code
• RAID 3 - bit-interleaved parity
• RAID 4 - Block-interleaved parity (independent
  access)
• RAID 5 - Block-interleaved parity (independent
  access)
• RAID 6 - P Q redundancy
• RAID 7 - heterogeneous devices

15
RAID (contd.)

• Unique contribution of the RAID is to address the
  need for redundancy.
• Principle redundancy compensate for reduced
  reliability
• Even though use of multiple heads and actuators
  result in higher transfer rates, they also
  introduce higher probability of failure.
• To compensate for this decreased reliability,
  RAID makes use of stored parity information to
  recover data lost due to disk failure.

16
RAID 1

• RAID 0 data is striped across many disks
  similar to interleaved memory.
• Not actually RAID.
• No redundancy but high transfer rate due to
  organization.
• RAID 1 redundancy is achieved by simple
  duplicating.
• Read can be performed by either disk which ever
  has lower seek and rotational time.
• Write can be performed simultaneous on the two
  disks.
• Recovery from failures is simple use the
  surviving disk.

17
RAID 2

• RAID 2 - RAID 5 have some kind of parity.
• RAID 2
• striping
• Hamming error correcting code
• 4 data disks , 3 Hamming code disks
• Effective in environment which are error prone.

18
RAID 3

• Similar to RAID but only one single redundant
  disk.
• Single redundant disk contains the parity info of
  the bits across the four disks.
• Single disk failure can be tolerated with
  slightly lower transfer rate until the failed
  disk is replaced.

19
RAID 4 and 5

• Independently accessible.
• Strip parity (text calls it block parity).
• Large strips.
• In RAID 4 parity is stored on a single disk, and
  in RAID 5 parity is distributed over disks.

20
RAID 6 and RAID 7

• RAID 6 real fault tolerance hardware included.
  Significantly more expensive than RAID 5.
• RAID 7 using distributed and heterogenous
  sources of data.

21
Summary

• Look at exercises 11.1, 11.2, 11.7, 11.8
• Hardware characteristics underlying IO devices
  are abstracted for presenting easier-to-use and
  uniform logical interface to the processes.
• Performance improvement is another important
  objective which is achieved through buffering
  schemes and scheduling algorithms.