Part 10: I/O System & Disk

• I/O Hardware
• I/O System Design and Implementation
– Design Objectives
– Kernel I/O Structure
– Kernel I/O Subsystem
– Performance Consideration
• Disk
– Disk Structure
– Disk Scheduling
!FCFS, SSTF, SCAN, C-SCAN, C-LOOK
– Disk Management
– Disk Reliability
Operating Systems 10.1 Part 10 I/O System & Disk
 I/O Hardware

Operating Systems 10.2 Part 10 I/O System & Disk

 I/O Hardware

• Devices vary in many dimensions

Operating Systems 10.3 Part 10 I/O System & Disk

 I/O System Design Objectives

• Efficiency
– Important because I/O operations often form a
bottleneck
– Most I/O devices are extremely slow compared with
main memory and the processor
– The area that has received the most attention is disk
I/O
• Generality
– Desirable to handle all devices in a uniform manner
– Applies to the way processes view I/O devices and
the way the operating system manages I/O devices
and operations

Operating Systems 10.4 Part 10 I/O System & Disk

 Kernel I/O Structure

Operating Systems 10.5 Part 10 I/O System & Disk

 Kernel I/O Subsystem

• I/O Scheduling
– Schedule I/O request through rearranging the order of
service, e.g. disk scheduling
– Can improve the system efficiency

Operating Systems 10.6 Part 10 I/O System & Disk

 Kernel I/O Subsystem

• Buffering
– Store data in memory
while transferring
between devices
– To cope with device
speed mismatch
– To cope with device
transfer size
mismatch, e.g. in
networking

Operating Systems 10.7 Part 10 I/O System & Disk

 Kernel I/O Subsystem

• Caching
– Fast memory holding copy of data
– To improve I/O efficiency for files that are
being written and reread rapidly.
• Spooling
– Hold output for a device
– Used in device that can serve only one
request at a time, e.g. printing

Operating Systems 10.8 Part 10 I/O System & Disk

 Performance Consideration

• Application I/O Interface

1. A process makes
an I/O request
and the request
is sent to the I/O
device where it is
processed.

2. The system call

1. A process makes an I/O returns after the
request and the request request is queued
is sent to the I/O device at I/O device.
where it is processed.
3. The requesting
2. The process is in the process continues
waiting state while the to execute its code
request is being while the request is
processed by I/O device. being processed by
I/O device.
3. The process is put back to
the ready queue when the 4. When the I/O
I/O request is completed. completes, the
requesting process
Blocking Asynchronous is signaled.
I/O I/O

Operating Systems 10.9 Part 10 I/O System & Disk

 Performance Consideration

• Using asynchronous I/O and double-buffer to

improve performance

b2

b1

Double User
Buffer Process

sử dụng 2 buffer để cho I/O device vừa ghi data vào b1 và process lấy data từ b2 ra

Operating Systems 10.10 Part 10 I/O System & Disk

 Performance Consideration
1 10

2
9

4 8

5 6

Operating Systems 10.11 Part 10 I/O System & Disk

 Performance Consideration

• I/O is a major factor in system performance

– It demands CPU to execute device driver
code
– Context switches due to interrupts may stress
CPU
– Data copying between I/O controllers and
physical memory also affect system
performance.

Operating Systems 10.12 Part 10 I/O System & Disk

 Part 7: I/O System & Disk

• I/O Hardware
• I/O System Design and Implementation
– Design Objectives
– Kernel I/O Structure
– Kernel I/O Subsystem
– Performance Consideration
• Disk
– Disk Structure
– Disk Scheduling
!FCFS, SSTF, SCAN, C-SCAN, C-LOOK
– Disk Management
– Disk Reliability
Operating Systems 10.13 Part 10 I/O System & Disk
 Disk Structure
Sectors
(1 sector = 512 bytes)
Disk heads

Platters

Operating Systems 10.14 Part 10 I/O System & Disk

 Disk Structure

• Disk drives are

addressed as large 1-
dimensional arrays of
physical blocks, where
the physical block is the
smallest unit of transfer
(usually 512 bytes)
• The array of physical
blocks is mapped into the
sectors of the disk
sequentially

Operating Systems 10.15 Part 10 I/O System & Disk

 Disk I/O

• To read a requested sector, disk drive needs to

position the disk head and perform data transfer.

• Position time has two major

components
-Seek time for the disk to move
the heads to the track containing
the desired sector (5 -25
milliseconds)
-Rotational latency for the disk to
rotate the desired sector to the
disk head (8 milliseconds)

Operating Systems 10.16 Part 10 I/O System & Disk

 Disk I/O (Cont.)

• How long to read or write N sectors?

• Positioning Time + Data Transfer Time (N)
• Positioning time: Seek Time + Rotational Latency
• Transfer time: (N / Number of Sectors per Track) ´ Full Rotation Time

• Some parameters:
• Seek time: 10 ms
• Full rotational time: 8 ms (Rotational latency: 4 ms)
• Sector size: 512 Bytes
• Track size: 1 K sectors per track

Operating Systems 10.17 Part 10 I/O System & Disk

 Disk I/O (Cont.)

• Random Request
– Seek Time + Rotational Latency + Transfer Time
10 ms + 4 ms + (1/1024) ´ 8ms = ~14 ms

• Time and bandwidth for random request of size:

– 4 KB = 8 sectors = 8 * 14 ms = 112 ms
– 128 KB = 256 sectors = 256 * 14 ms = 3.5 secs
– 1 MB = 2048 sectors = 2048 * 14 ms = 28 seconds !!

Operating Systems 10.18 Part 10 I/O System & Disk

 Disk I/O (Cont.)
• Sequential Request
– 4KB? 8 sectors
! 1 positioning time = 14 ms
! Transfer time = (8 / 1024) ´ full rotational time = 0.0625 ms
! Total = ~14 ms

– 128 KB? 256 sectors

! Still fits in a track, hence 1 positioning time = 14 ms
! Transfer time = (256 / 1024) ´ full rotational time = 2 ms
! Total = 16 ms

– 1 MB? 2048 sectors

! Does not fit in a track, fits in 2 tracks (at best) or 3 tracks (at worst).
! Read one track: positioning time + full rotation time = 14 ms + 8 ms = 22 ms
! Read two tracks: 2 * 22 ms = 44 ms (vs. 28 seconds of random I/Os)

Operating Systems 10.19 Part 10 I/O System & Disk

 Disk Scheduling

• The OS is responsible for using hardware

efficiently — for disk drives this means having a
fast access time and high disk bandwidth
• Minimize seek time
• Seek time » seek distance
• Disk bandwidth is the total number of bytes
transferred, divided by the total time between the
1st request and completion of the last transfer

Operating Systems 10.20 Part 10 I/O System & Disk

 Disk Scheduling (Cont.)

• Several algorithms exist to schedule the

servicing of disk I/O requests
• We illustrate them with a request queue (0 -199)
98, 183, 37, 122, 14, 124, 65, 67
• Assume that head is initially at cylinder 53

Operating Systems 10.21 Part 10 I/O System & Disk

 First-Come-First-Served (FCFS)

Operating Systems 10.22 Part 10 I/O System & Disk

 FCFS (Cont.)

Total head movement =

|98 - 53| + |183 - 98| + |37 - 183| + |122 - 37| +
|14 -122| + |124 -14| + |65 -124| + |67 - 65|
= 640

Problem: FCFS may have wild swings back and forth!

Operating Systems 10.23 Part 10 I/O System & Disk

 Shortest-Seek-Time-First (SSTF)

• Selects the request with the minimum seek time

from the current head position
• SSTF scheduling is a form of SJF scheduling;
may have starvation under heavy load, since
distant requests may never be serviced
• Note: requests are from processes executing
concurrently

Operating Systems 10.24 Part 10 I/O System & Disk

 SSTF (Cont.)

Operating Systems 10.25 Part 10 I/O System & Disk

 SSTF (Cont.)

Total head movement =

|65 - 53| + |67 - 65| + |37 - 67| + |14 - 37| +

|98 -14| + |122 - 98| + |124 - 122| + |183 - 124|

= 236

Problem: SSTF may cause starvation!

Operating Systems 10.26 Part 10 I/O System & Disk

 SCAN

• Disk arm starts at one end of the disk, and

moves toward the other end, servicing requests
until it gets to the other end of the disk, where
the head movement is reversed and servicing
continues
• Also called the elevator algorithm
• Assuming here the head is moving toward 0
• Starvation unlikely but servicing requests may
be delayed

Operating Systems 10.27 Part 10 I/O System & Disk

 SCAN (Cont.)

Operating Systems 10.28 Part 10 I/O System & Disk

 SCAN (Cont.)

Total head movement =

|37 - 53| + |14 - 37| + |0 - 14| + |65 - 0| + |67 - 65|

+ |98 - 67| + |122 - 98| + |124 - 122| + |183 - 124|

= 236

After the head reverses direction, does it needs to

serve requests in the reversed direction immediately?

Operating Systems 10.29 Part 10 I/O System & Disk

 Circular-Scan (C-SCAN)

• Head moves from low end to high end of the

disk, servicing requests along
• When reaching the high end, head immediately
returns to the beginning of the disk, without
servicing any requests

Operating Systems 10.30 Part 10 I/O System & Disk

 C-SCAN (Cont.)

Operating Systems 10.31 Part 10 I/O System & Disk

 C-SCAN (Cont.)
Total head movement =
|65 - 53| + |67 - 65| + |98 - 67| + |122 - 98| +
|124 - 122| + |183 -124| + |199 - 183| +
|0 -199| + |14 - 0| + |37 -14|
= 382

Operating Systems 10.32 Part 10 I/O System & Disk

 C-LOOK

• Version of C-SCAN
• Head moves from low end to high end of the
disk, servicing requests along
• When reaching the last request, head returns
immediately to the lowest cylinder # request
• From there it services the requests towards the
high end of the disk

Chỉ chạy mũi tên trong max và min của cái queue thôi

Operating Systems 10.33 Part 10 I/O System & Disk

 C-LOOK (Cont.)

Operating Systems 10.34 Part 10 I/O System & Disk

 C-LOOK (Cont.)

Total head movement =

|65 - 53| + |67 - 65| + |98 - 67| + |122 - 98| +

|124 -122| + |183 -124| + |14 -183| + |37 - 14|

= 322

Does SCAN, C-SCAN and C-LOOK really

solve the starvation problem?

Operating Systems 10.35 Part 10 I/O System & Disk

 Selecting a Disk-Scheduling Algorithm

• SSTF is common and has a natural appeal

• SCAN and C-SCAN (or LOOK and C-LOOK)
perform better for systems that place a heavy
load on the disk (since starvation is unlikely)
• Performance depends on the number and types
of requests

Operating Systems 10.36 Part 10 I/O System & Disk

 Selecting a Disk-Scheduling Algorithm (Cont.)

• Requests for disk service can be influenced by

the file-allocation method:
– A program reading a contiguously allocated
file will generate requests that are close
together
– A linked or indexed file, may generate
requests that are widely apart, resulting in
greater head movement

Operating Systems 10.37 Part 10 I/O System & Disk

 Disk Management

• Low-level formatting (physical formatting) —

Dividing a disk into sectors that the disk
controller can read and write
• To use a disk to hold files, the OS needs to
record its own data structures on the disk.
Two steps:
– Partition the disk into one or more groups of
cylinders
– Logical formatting or “making a file system”

Operating Systems 10.38 Part 10 I/O System & Disk

 Disk Management (Cont.)

Operating Systems 10.39 Part 10 I/O System & Disk

 Disk Reliability

• Several improvements in disk-use techniques

involve multiple disks working cooperatively
• Disk striping uses a group of disks as one
storage unit
– Each block is broken into several sub-blocks,
with one sub-block stored on each disk
– Time to transfer a block into memory is faster
because all sub-blocks are transferred in
parallel

triping là bắn bắn

Operating Systems 10.40 Part 10 I/O System & Disk

 Disk Reliability (cont.)
Block 0
Block 4
Block 8

Block 1
Block 5
Block 10

Block 2
Block 6

Block 3
Block 7

with n drives, block i of a file stores on drive (i mod n) + 1

Operating Systems 10.41 Part 10 I/O System & Disk
 Disk Reliability (cont.)

• Mirroring keeps duplicate of each disk

– A logical disk consists of two physical disks
– If one fails, the data can be read from the
other

Operating Systems 10.42 Part 10 I/O System & Disk

 Disk Reliability (cont.)

• RAID (Redundant-Array-of-Independent-Disks)
improves performance and reliability of the
storage system by storing redundant data
– RAID 0: non-redundant striping
– Mirroring (RAID 1) keeps duplicate of each
disk
– Mirror of Strip (RAID 0+1) or Strip of Mirrors
(RAID 1+0) provides high performance and
high reliability

Operating Systems 10.43 Part 10 I/O System & Disk

 Disk Reliability (cont.)

Operating Systems 10.44 Part 10 I/O System & Disk

 Summary

• Device-driver layer hides the differences among

device controllers from the I/O subsystem of the
kernel.
• The I/O subsystem provides functions for: I/O
scheduling, buffering, spooling, error handling,
device reservation, and device name translation.
• I/O system calls are costly in terms of CPU
consumption, because of the many layers of
software between a physical device and the
application.

Operating Systems 10.45 Part 10 I/O System & Disk

 Summary

• Disk scheduling algorithms optimize seek time:

– FCFS, SSTF, SCAN, C-SCAN, C-LOOK
• Disk management requires disk partitioning and
formatting
• Disk striping improve performance
• Mirroring improves reliability

Operating Systems 10.46 Part 10 I/O System & Disk