Cs8493 unit 4

CS8493 OPERATING SYSTEMS
UNIT – IV
Dr.A.Kathirvel, Professor &Head, Computer Science & Engg.
M N M Jain Engineering College, Chennai
1

Unit - IV
FILE SYSTEMS AND I/O SYSTEMS
Mass Storage system – Overview of Mass Storage Structure, Disk
Structure, Disk Scheduling and Management, swap space
management; File-System Interface – File concept, Access methods,
Directory Structure, Directory organization, File system mounting,
File Sharing and Protection; File System Implementation- File
System Structure, Directory implementation, Allocation Methods,
Free Space Management, Efficiency and Performance, Recovery; I/O
Systems – I/O Hardware, Application I/O interface, Kernel I/O
subsystem, Streams, Performance.
Abraham Silberschatz, Peter Baer Galvin and Greg Gagne, “Operating System
Concepts”, 9 th Edition, John Wiley and Sons Inc., 2012.
2

Overview of Mass Storage Structure
 Magnetic disks provide bulk of secondary storage of modern computers
Drives rotate at 60 to 250 times per second
Transfer rate is rate at which data flow between drive and computer
Positioning time (random-access time) is time to move disk arm to
desired cylinder (seek time) and time for desired sector to rotate
under the disk head (rotational latency)
Head crash results from disk head making contact with the disk
surface -- That’s bad
 Disks can be removable
 Drive attached to computer via I/O bus
Busses vary, including EIDE, ATA, SATA, USB, Fibre Channel, SCSI,
SAS, Firewire
Host controller in computer uses bus to talk to disk controller built
into drive or storage array
3

Hard Disks
 Platters range from .85 ” to 14 ”
(historically)
Commonly 3.5”, 2.5”, and 1.8”
 Range from 30GB to 3TB per drive
 Performance
Transfer Rate – theoretical – 6 Gb/sec
Effective Transfer Rate – real –
1Gb/sec
Seek time from 3ms to 12ms – 9ms
common for desktop drives
Average seek time measured or
calculated based on 1/3 of tracks
Latency based on spindle speed
1 / (RPM / 60) = 60 / RPM
Average latency = ½ latency 4

Hard Disk Performance
 Access Latency = Average access time = average seek time + average
latency
For fastest disk 3ms + 2ms = 5ms
For slow disk 9ms + 5.56ms = 14.56ms
 Average I/O time = average access time + (amount to transfer /
transfer rate) + controller overhead
 For example to transfer a 4KB block on a 7200 RPM disk with a 5ms
average seek time, 1Gb/sec transfer rate with a .1ms controller
overhead =
5ms + 4.17ms + 0.1ms + transfer time =
Transfer time = 4KB / 1Gb/s * 8Gb / GB * 1GB / 10242KB = 32 /
(10242) = 0.031 ms
Average I/O time for 4KB block = 9.27ms + .031ms = 9.301ms
5

The First Commercial Disk Drive
1956
IBM RAMDAC computer
included the IBM Model 350
disk storage system
5M (7 bit) characters
50 x 24” platters
Access time = < 1 second
6

Solid-State Disks
Nonvolatile memory used like a hard drive
Many technology variations
Can be more reliable than HDDs
More expensive per MB
Maybe have shorter life span
Less capacity
But much faster
Busses can be too slow -> connect directly to PCI for
example
No moving parts, so no seek time or rotational latency
7

Magnetic Tape
 Was early secondary-storage medium
Evolved from open spools to cartridges
 Relatively permanent and holds large quantities of data
 Access time slow
 Random access ~1000 times slower than disk
 Mainly used for backup, storage of infrequently-used data,
transfer medium between systems
 Kept in spool and wound or rewound past read-write head
 Once data under head, transfer rates comparable to disk
140MB/sec and greater
 200GB to 1.5TB typical storage
 Common technologies are LTO-{3,4,5} and T10000
8

Disk Structure
 Disk drives are addressed as large 1-dimensional arrays of logical
blocks, where the logical block is the smallest unit of transfer
Low-level formatting creates logical blocks on physical media
 The 1-dimensional array of logical blocks is mapped into the sectors
of the disk sequentially
Sector 0 is the first sector of the first track on the outermost
cylinder
Mapping proceeds in order through that track, then the rest of
the tracks in that cylinder, and then through the rest of the
cylinders from outermost to innermost
Logical to physical address should be easy
Except for bad sectors
Non-constant # of sectors per track via constant angular
velocity
9

Disk Attachment
Host-attached storage accessed through I/O ports talking to
I/O busses
SCSI itself is a bus, up to 16 devices on one cable, SCSI
initiator requests operation and SCSI targets perform tasks
Each target can have up to 8 logical units (disks attached
to device controller)
FC is high-speed serial architecture
Can be switched fabric with 24-bit address space – the
basis of storage area networks (SANs) in which many
hosts attach to many storage units
I/O directed to bus ID, device ID, logical unit (LUN)
10

Storage Array
Can just attach disks, or arrays of disks
Storage Array has controller(s), provides features to attached
host(s)
Ports to connect hosts to array
Memory, controlling software (sometimes NVRAM, etc)
A few to thousands of disks
RAID, hot spares, hot swap (discussed later)
Shared storage -> more efficiency
Features found in some file systems
Snaphots, clones, thin provisioning, replication,
deduplication, etc
11

Storage Area Network
Common in large storage environments
Multiple hosts attached to multiple storage arrays
– flexible
SAN is one or more storage arrays
Connected to one or more Fibre Channel switches
12

Storage Area Network
Hosts also attach to the switches
Storage made available via LUN Masking from
specific arrays to specific servers
Easy to add or remove storage, add new host and
allocate it storage
Over low-latency Fibre Channel fabric
Why have separate storage networks and
communications networks?
Consider iSCSI, FCOE
13

Network-Attached Storage
Network-attached storage (NAS) is storage made available
over a network rather than over a local connection (such
as a bus)
Remotely attaching to file systems
NFS and CIFS are common protocols
Implemented via remote procedure calls (RPCs) between
host and storage over typically TCP or UDP on IP network
iSCSI protocol uses IP network to carry the SCSI protocol
Remotely attaching to devices (blocks)
14

Disk Scheduling
The operating system is responsible for using hardware
efficiently — for the disk drives, this means having a fast
access time and 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 first request for
service and the completion of the last transfer
There are many sources of disk I/O request
OS
System processes
Users processes 15

Disk Scheduling
I/O request includes input or output mode, disk address,
memory address, number of sectors to transfer
OS maintains queue of requests, per disk or device
Idle disk can immediately work on I/O request, busy disk
means work must queue
Optimization algorithms only make sense when a
queue exists
Note that drive controllers have small buffers and can
manage a queue of I/O requests (of varying “depth”)
Several algorithms exist to schedule the servicing of disk
I/O requests
16

FCFS Disk Scheduling
The analysis is true for one or many platters
We illustrate scheduling algorithms with a request queue
(0-199)
98, 183, 37, 122,
14, 124, 65, 67
Head pointer 53
17

SSTF
Shortest Seek Time First
selects the request with
the minimum seek time
from the current head
position
SSTF scheduling is a form
of SJF scheduling; may
cause starvation of some
requests
Illustration shows total
head movement of 236
cylinders
18

SCAN
The 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.
SCAN algorithm
Sometimes called the
elevatoralgorithm
Illustration shows total
head movement of 208
cylinders
But note that if requests
are uniformly dense,
largest density at other
end of disk and those wait
the longest 19

C-SCAN
Provides a more uniform wait time than SCAN
The head moves from one end of the disk to the other, servicing
requests as it goes
When it reaches the other end, however, it immediately
returns to the beginning of the disk, without servicing any
requests on the return trip
Treats the cylinders as a
circular list that wraps around
from the last cylinder to the
first one
Total number of cylinders?
20

C-LOOK
LOOK a version of SCAN, C-LOOK a version of C-
SCAN
Arm only goes as far
as the last request in
each direction, then
reverses direction
immediately, without
first going all the way
to the end of the disk
Total number of
cylinders?
21

Selecting a Disk-Scheduling Algorithm
 SSTF is common and has a natural appeal
 SCAN and C-SCAN perform better for systems that place a heavy load
on the disk - Less starvation
 Performance depends on the number and types of requests
 Requests for disk service can be influenced by the file-allocation
method and metadata layout
 The disk-scheduling algorithm should be written as a separate
module of the operating system, allowing it to be replaced with a
different algorithm if necessary
 Either SSTF or LOOK is a reasonable choice for the default algorithm
 What about rotational latency?
Difficult for OS to calculate
 How does disk-based queueing effect OS queue ordering efforts?
22

Disk Management
 Low-level formatting, or physical formatting — Dividing a disk into
sectors that the disk controller can read and write
Each sector can hold header information, plus data, plus error
correction code (ECC)
Usually 512 bytes of data but can be selectable
 To use a disk to hold files, the operating system still needs to record
its own data structures on the disk
Partition the disk into one or more groups of cylinders, each
treated as a logical disk
Logical formatting or “making a file system”
To increase efficiency most file systems group blocks into clusters
Disk I/O done in blocks
File I/O done in clusters
23

Disk Management
 Raw disk access for apps that
want to do their own block
management, keep OS out of
the way (databases for
example)
 Boot block initializes system
The bootstrap is stored in
ROM
Bootstrap loader program
stored in boot blocks of
boot partition
 Methods such as sector
sparing used to handle bad
blocks
Booting from a Disk in Windows
24

Swap-Space Management
Swap-space — Virtual memory uses disk space as an
extension of main memory
Less common now due to memory capacity increases
Swap-space can be carved out of the normal file system,
or, more commonly, it can be in a separate disk partition
(raw)
Swap-space management
4.3BSD allocates swap space when process starts;
holds text segment (the program) and data segment
Kernel uses swap maps to track swap-space use
25

Data Structures for
Swapping on Linux Systems
Solaris 2 allocates swap space only when a dirty page is forced out of
physical memory, not when the virtual memory page is first created
File data written to swap space until write to file system requested
Other dirty pages go to swap space due to no other home
Text segment pages thrown out and reread from the file system as
needed
What if a system runs out of swap space?
Some systems allow multiple swap spaces
Swap-Space Management
26

RAID Structure
 RAID – redundant array of inexpensive disks
multiple disk drives provides reliability via redundancy
 Increases the mean time to failure
 Mean time to repair – exposure time when another failure could
cause data loss
 Mean time to data loss based on above factors
 If mirrored disks fail independently, consider disk with 1300,000
mean time to failure and 10 hour mean time to repair
Mean time to data loss is 100, 0002 / (2 ∗ 10) = 500 ∗ 106 hours,
or 57,000 years!
 Frequently combined with NVRAM to improve write performance
 Several improvements in disk-use techniques involve the use of
multiple disks working cooperatively
27

RAID
 Disk striping uses a group of disks as one storage unit
 RAID is arranged into six different levels
 RAID schemes improve performance and improve the reliability of
the storage system by storing redundant data
Mirroring or shadowing (RAID 1) keeps duplicate of each disk
Striped mirrors (RAID 1+0) or mirrored stripes (RAID 0+1)
provides high performance and high reliability
Block interleaved parity (RAID 4, 5, 6) uses much less redundancy
 RAID within a storage array can still fail if the array fails, so automatic
replication of the data between arrays is common
 Frequently, a small number of hot-spare disks are left unallocated,
automatically replacing a failed disk and having data rebuilt onto
them
28

RAID Levels RAID (0 + 1) and (1 + 0)
29

Other Features
Regardless of where RAID implemented, other useful
features can be added
Snapshot is a view of file system before a set of changes
take place (i.e. at a point in time)
Replication is automatic duplication of writes between
separate sites
For redundancy and disaster recovery
Can be synchronous or asynchronous
Hot spare disk is unused, automatically used by RAID
production if a disk fails to replace the failed disk and
rebuild the RAID set if possible
Decreases mean time to repair
30

Extensions
RAID alone does not prevent or detect data corruption or
other errors, just disk failures
Solaris ZFS adds checksums of all data and metadata
Checksums kept with pointer to object, to detect if object
is the right one and whether it changed
Can detect and correct data and metadata corruption
ZFS also removes volumes, partitions
Disks allocated in pools
Filesystems with a pool share that pool, use and
release space like malloc() and free() memory
allocate / release calls
31

ZFS Checksums All Metadata & Data
Traditional and Pooled Storage
32

Stable-Storage Implementation
 Write-ahead log scheme requires stable storage
 Stable storage means data is never lost (due to failure, etc)
 To implement stable storage:
Replicate information on more than one nonvolatile storage media with
independent failure modes
Update information in a controlled manner to ensure that we can
recover the stable data after any failure during data transfer or recovery
 Disk write has 1 of 3 outcomes
Successful completion - The data were written correctly on disk
Partial failure - A failure occurred in the midst of transfer, so only some
of the sectors were written with the new data, and the sector being
written during the failure may have been corrupted
Total failure - The failure occurred before the disk write started, so the
previous data values on the disk remain intact
33

Stable-Storage Implementation
If failure occurs during block write, recovery procedure
restores block to consistent state
System maintains 2 physical blocks per logical block
and does the following:
Write to 1st physical
When successful, write to 2nd physical
Declare complete only after second write
completes successfully
Systems frequently use NVRAM as one physical to
accelerate
34

File Concept
Contiguous logical address space
Types:
Data
numeric
character
binary
Program
Contents defined by file’s creator
Many types
Consider text file, source file,
executable file File info Window on Mac OS X
35

File Attributes
 Name – only information kept in human-readable form
 Identifier – unique tag (number) identifies file within file system
 Type – needed for systems that support different types
 Location – pointer to file location on device
 Size – current file size
 Protection – controls who can do reading, writing, executing
 Time, date, and user identification – data for protection, security,
and usage monitoring
 Information about files are kept in the directory structure, which is
maintained on the disk
 Many variations, including extended file attributes such as file
checksum
 Information kept in the directory structure
36

File Operations
File is an abstract data type
Create
Write – at write pointer location
Read – at read pointer location
Reposition within file - seek
Delete
Truncate
Open(Fi) – search the directory structure on disk for entry
Fi, and move the content of entry to memory
Close (Fi) – move the content of entry Fi in memory to
directory structure on disk
37

Open Files
Several pieces of data are needed to manage open
files:
Open-file table: tracks open files
File pointer: pointer to last read/write location, per
process that has the file open
File-open count: counter of number of times a file is
open – to allow removal of data from open-file table
when last processes closes it
Disk location of the file: cache of data access
information
Access rights: per-process access mode information
38

Open File Locking
Provided by some operating systems and file systems
Similar to reader-writer locks
Shared lock similar to reader lock – several processes
can acquire concurrently
Exclusive lock similar to writer lock
Mediates access to a file
Mandatory or advisory:
Mandatory – access is denied depending on locks held
and requested
Advisory – processes can find status of locks and decide
what to do
39

File Locking Example – Java API
import java.io.*;
import java.nio.channels.*;
public class LockingExample {
public static final boolean EXCLUSIVE=false;
public static final boolean SHARED = true;
public static void main(String arsg[]) throws IOException {
FileLock sharedLock =null;
FileLock exclusiveLock = null;
try {
RandomAccessFileraf = new RandomAccessFile("file.txt", "rw");
// get the channel for the file
FileChannel ch = raf.getChannel();
// this locks the firsthalf of the file - exclusive
exclusiveLock = ch.lock(0, raf.length()/2, EXCLUSIVE);
/** Now modify the data . . . */
// release the lock
exclusiveLock.release();
// this locks the second half of the file -
shared
sharedLock =ch.lock(raf.length()/2+1,
raf.length(), SHARED);
/** Now read the data . . . */
// release the lock
sharedLock.release();
} catch (java.io.IOExceptionioe) {
System.err.println(ioe);
}finally {
if (exclusiveLock != null)
exclusiveLock.release();
if (sharedLock !=null)
sharedLock.release();
}
}
}
40

File Types:
Name, Extension
Sequential:
access File
41

File Structure
 None - sequence of words, bytes
 Simple record structure
Lines
Fixed length
Variable length
 Complex Structures
Formatted document
Relocatable load file
 Can simulate last two with first method by inserting appropriate
control characters
 Who decides:
Operating system
Program
42

Access Methods
 Sequential Access
read next
write next
reset
no read after last write
(rewrite)
 Direct Access – file is fixed length logical records
read n
write n
position to n
read next
write next
rewrite n n = relative block number
 Relative block numbers allow OS to decide where file should be
placed
43

Other Access Methods
Can be built on top of base methods
General involve creation of an index for the file
Keep index in memory for fast determination of location of
data to be operated on (consider UPC code plus record of
data about that item)
If too large, index (in memory) of the index (on disk)
IBM indexed sequential-access method (ISAM)
Small master index, points to disk blocks of secondary
index
File kept sorted on a defined key
All done by the OS
VMS operating system provides index and relative files as
another example (see next slide)
44

Example of Index and Relative Files
Simulation of Sequential
Access on Direct-access File
45

Directory Structure
A collection of nodes containing information about all files
F 1 F 2
F 3
F 4
F n
Directory
Files
Both the directory structure
and the files reside on disk
A Typical File-system Organization
46

Disk Structure
Disk can be subdivided into partitions
Disks or partitions can be RAID protected against failure
Disk or partition can be used raw – without a file system, or
formatted with a file system
Partitions also known as minidisks, slices
Entity containing file system known as a volume
Each volume containing file system also tracks that file
system’s info in device directory or volume table of
contents
As well as general-purpose file systems there are many
special-purpose file systems, frequently all within the
same operating system or computer
47

Types of File Systems
 We mostly talk of general-purpose file systems
 But systems frequently have may file systems, some general- and
some special- purpose
 Consider Solaris has
tmpfs – memory-based volatile FS for fast, temporary I/O
objfs – interface into kernel memory to get kernel symbols for
debugging
ctfs – contract file system for managing daemons
lofs – loopback file system allows one FS to be accessed in place
of another
procfs – kernel interface to process structures
ufs, zfs – general purpose file systems
48

Operations Performed on Directory
Search for a file
Create a file
Delete a file
List a directory
Rename a file
Traverse the file system
49

Directory Organization
Efficiency – locating a file quickly
Naming – convenient to users
Two users can have same name for different files
The same file can have several different names
Grouping – logical grouping of files by properties, (e.g., all
Java programs, all games, …)
The directory is organized logically to obtain
Single-Level Directory
 A single directory for all users
 Naming problem
 Grouping problem
50

Two-Level Directory
Separate directory for each user
 Path name
 Can have the same file
name for different user
 Efficient searching
 No grouping capability
Tree-Structured
 Efficient searching
 Grouping Capability
 Current directory
(working directory)
cd /spell/mail/prog
type list
51

Tree-Structured Directories (Cont)
Absolute or relative path name
Creating a new file is done in current directory
Delete a file rm <file-name>
Creating a new subdirectory is done in current
directory mkdir <dir-name>
Example: if in current directory /mail
mkdir count
Deleting “mail”  deleting the entire subtree rooted by “mail”
52

Acyclic-Graph Directories
Have shared subdirectories and files
 Two different names (aliasing)
 If dict deletes list  dangling pointer
 Solutions:
Backpointers, so we can delete all
pointers
Variable size records a problem
Backpointers using a daisy chain
organization
Entry-hold-count solution
 New directory entry type
Link – another name (pointer) to an
existing file
Resolve the link – follow pointer to
locate the file 53

General Graph Directory
 How do we guarantee no
cycles?
Allow only links to file
not subdirectories
Garbage collection
Every time a new link
is added use a cycle
detection algorithm to
determine whether it
is OK
54

File System Mounting
A file system must be mounted before it can be accessed
A unmounted file system (Fig is mounted at a mount point)
Mount Point
55

File Sharing
Sharing of files on multi-user systems is desirable
Sharing may be done through a protection scheme
On distributed systems,files may be shared across a network
Network File System (NFS) is a common distributed file-sharing
method
If multi-user system
User IDs identify users, allowing permissions and protections
to be per-user
Group IDs allow users to be in groups, permitting group
access rights
Owner of a file / directory
Group of a file / directory
56

File Sharing – Remote File Systems
 Uses networking to allow file system access between systems
Manually via programs like FTP
Automatically, seamlessly using distributed file systems(remote
directory are visible from a local machine)
Semi automatically via the world wide web(browser needed to access
to the remote file used to transfer file)
 Client-server model allows clients to mount remote file systems from
servers
Server can serve multiple clients
Client and user-on-client identification is insecure or complicated
NFS is standard UNIX client-server file sharing protocol
CIFS (common internet file system) is standard Windows protocol
Standard operating system file calls are translated into remote calls
 Distributed Information Systems (distributed naming services) such as
LDAP, DNS, NIS, Active Directory implement unified access to information
needed for remote computing 57

File Sharing – Failure Modes
All file systems have failure modes
For example corruption of directory structures or other
non-user data, called metadata
Remote file systems add new failure modes, due to
network failure, server failure (remote file system
protocol)
Recovery from failure can involve state information about
status of each remote request
Stateless protocols such as NFS v3 include all information
in each request, allowing easy recovery but less security
Both server and client maintain the current activities
(nfs protocol)
58

File Sharing – Consistency Semantics
 It evaluating any file system that support file sharing.
 Specify how multiple users are to access a shared file simultaneously
Similar to Ch 5 process synchronization algorithms
Tend to be less complex due to disk I/O and network latency (for
remote file systems
Andrew File System (AFS) implemented complex remote file sharing
semantics
Unix file system (UFS) implements:
Writes to an open file visible immediately to other users of the
same open file
Sharing file pointer to allow multiple users to read and write
concurrently
Andrew File system has session semantics
Writes only visible to sessions starting after the file is closed
59

Protection
File owner/creator should be able to control:
what can be done
by whom
Types of access
Read
Write
Execute
Append
Delete
List
Windows 7 Access-Control
List Management
60

Access Lists and Groups
 Mode of access: read, write, execute
 Three classes of users on Unix / Linux
RWX
( user who created the file) a) owner access 7  111
RWX
(who sharing the file) b) group access 6  1 1 0
RWX
(all other user ) c) public access 1  0 0 1
 Ask manager to create a group (unique name), say G, and add some users to the
group.For a particular file (say game) or subdirectory, define an appropriate
access.
 Disk provide of secondary storage on which file system are maintained.
 2 characteristics
 Disk can be rewrite in place,-*read modify and write block into the same
place
 Disk can access directly any block of information it contain – (file either
sequential or random one file moving to read- write head for the disk to
rotate.
Attach a group to a file: chgrp G game
61

A Sample UNIX Directory Listing
 It provide efficient and convenient access to disk by allowing data to be stored,
located and retrieved
 It poses 2 problem
How file system should look for user(defining file attribute, opeartion,disk
structure for organizing file)
Creating algorithm and data structure to map logical to physical address.
62

File-System Structure
 File structure
Logical storage unit
Collection of related information
 File system resides on secondary storage (disks)
Provided user interface to storage, mapping logical to physical
Provides efficient and convenient access to disk by allowing data
to be stored, located retrieved easily
 Disk provides in-place rewrite and random access
I/O transfers performed in blocks of sectors (usually 512 bytes)
 File control block – storage structure consisting of information about
a file(permission, location and ownership)
 Device driver controls the physical device
 File system organized into layers

File System Layers
 Device drivers manage I/O devices at the I/O control layer
Given commands like “read drive1, cylinder 72, track 2, sector 10,
into memory location 1060” outputs low-level hardware specific
commands to hardware controller
 Basic file system given command like “retrieve block 123” translates
to device driver
 Also manages memory buffers and caches (allocation, freeing,
replacement)
Buffers hold data in transit
Caches hold frequently used data
 File organization module understands files, logical address, and
physical blocks
 Translates logical block # to physical block #
 Manages free space, disk allocation 65

File System Layers
 Logical file system manages metadata information
Translates file name into file number, file handle, location by
maintaining file control blocks (inodes in UNIX)
Directory management
Protection
 Layering useful for reducing complexity and redundancy, but adds
overhead and can decrease performanceTranslates file name into file
number, file handle, location by maintaining file control blocks (inodes
in UNIX)
Logical layers can be implemented by any coding method according
to OS designer
 Use of layers
Including the decision about
how many layer
Which layer should use 66

Directory Implementation
 Selectionof directory allocationand directory management algorithmaffect the
performance and reliability of file system
 Linear list of file names with pointer to the data blocks
 Simple to program
 Time-consuming to execute
Linear search time
Could keep ordered alphabetically via linked list or use B+ tree
ADVANTAGE
 it also be used to decrease the time required to delete file
 we can mark blank space and attached to free list
Disadvantage
 finding a file requires a linear search
 Access is very slow
67

Directory Implementation
Hash Table – linear list with hash data structure
It will return the pointer to the file name
Decreases directory search time
Collisions – situations where two file names hash to
the same location
Drawback
Only good if entries are fixed size, or use chained-
overflow method
68

Allocation Methods - Contiguous
 How to allocate space to these files in disk space
 An allocation method refers to how disk blocks are allocated for files:
 Contiguous allocation – each file occupies set of contiguous blocks
Best performance in most cases
Simple – only starting location (block #) and length (number of
blocks) are required
Problems include finding space for file, knowing file size, external
fragmentation, need for compaction off-line (downtime) or on-
line
One difficult is finding space for a new file
Dynamic storage allocation problem
It satisfy a request size n from list of free holes(fist / best fit)
Storage utilization increase , but fist fit faster
69

Contiguous Allocation
• Mapping from
logical to physical
LA/512
Q
R
Block to be accessed = Q +
starting address
Displacement into block = R
Preventing
copy an entire file system onto
another disk – (Compaction)
70

Extent-Based Systems
Many newer file systems (i.e., Veritas File System) use a
modified contiguous allocation scheme
Extent-based file systems allocate disk blocks in extents
If file not to be large enough, another chunck of
contiguous space known as extent
Other file system owner set extent size , but it also
inefficient if it incorrect
An extent is a contiguous block of disks
Extents are allocated for file allocation
A file consists of one or more extents
71

Allocation Methods - Linked
 Linked allocation– each file a linked list of blocks
 File ends at nil pointer
 No external fragmentation
 Each block contains pointer to next block
 No compaction, external fragmentation
 Free space management system called when new block needed
 Improve efficiency by clustering blocks into groups but increases internal
fragmentation
 Reliability can be a problem
 Locating a block can take many I/Os and disk seeks
 Size of the file need not to be declared when the file is created.
 Drawback
 It can be used effectively only for sequential access file
 If find ith block – start from beginning of the file and follow the pointer until
we get ith block. Space required for pointer
72

Allocation Methods – Linked
Solution:1.cluster – to collect block into multiples 2. FAT – unused block -0
 FAT (File Allocation Table) variation
 Beginning of volume has table, indexed by block number
 Much like a linked list, but faster on disk and cacheable
 New block allocation simple
 Each file is a linked list of disk blocks: blocks may be scattered anywhere on the
disk
pointerblock =  Mapping
LA/511
Q
R
Block to be accessed is the Qth block in the linked chain of blocks representing file.
Displacement into block = R + 1
73

Linked Allocation
File-Allocation Table
74

Allocation Methods - Indexed
 Indexed allocation
Each file has its own index block(s) of pointers to its data blocks
 Logical view
index table
Example
75

Indexed Allocation
 Need index table
 Random access
 Dynamic access without external fragmentation, but have overhead
of index block
 Mapping from logical to physical in a file of maximum size of 256K
bytes and block size of 512 bytes. We need only 1 block for index
table
LA/512
Q
R
Q = displacement into index table
R = displacement into block
 Mapping from logical to physical in a file of unbounded length (block
size of 512 words)
 Linked scheme – Link blocks of index table (no limit on size)
 Two-level index (4K blocks could store 1,024 four-byte pointers in
outer index -> 1,048,567 data blocks and file size of up to 4GB)
LA / (512 x 511)
Q1
R1
R1 / 512
Q2
R2
LA / (512 x 512)
Q1
R1
R1 / 512
Q2
R2
76

Indexed Allocation – Mapping
77

Combined Scheme: UNIX UFS
More index blocks than can be addressed with 32-bit file pointer
4K bytes per block, 32-bit addresses
78

Performance
 Best method depends on file access type
 Contiguous great for sequential and random
 Linked good for sequential, not random
 Declare access type at creation -> select either contiguous or linked
 Indexed more complex
 Single block access could require 2 index block reads then data block read
 Clustering can help improve throughput, reduce CPU overhead
 Adding instructions to the execution path to save one disk I/O is reasonable
 Intel Core i7 Extreme Edition 990x (2011)at 3.46Ghz = 159,000MIPS
http://en.wikipedia.org/wiki/Instructions_per_second
 Typical disk drive at 250 I/Os per second
159,000 MIPS/ 250 = 630 million instructions during one disk I/O
 Fast SSD drives provide 60,000 IOPS
159,000 MIPS/ 60,000 = 2.65 millions instructions during one disk I/O
79

Free-Space Management
 File system maintains free-space list to track available
blocks/clusters
(Using term “block” for simplicity)
 Bit vector or bit map (n blocks)
…
0 1 2 n-1
bit[i] =

1  block[i] free
0  block[i] occupied
Block number calculation
(number of bits per word) *
(number of 0-value words) +
offset of first 1 bit
CPUs have instructions to return offset within word of first “1” bit
 Bit map requires extra space. Example:
block size = 4KB = 212 bytes disk size = 240 bytes (1 terabyte)
n = 240/212 = 228 bits (or 32MB)
if clusters of 4 blocks -> 8MB of memory
 Easy to get contiguous files
80

Linked Free Space List on Disk
 Linked list (free list)
 Cannot get contiguous
space easily
 No waste of space
 No need to traverse the
entire list (if # free
blocks recorded)
81

Grouping
Modify linked list to store address of next n-1 free
blocks in first free block, plus a pointer to next block
that contains free-block-pointers (like this one)
Counting
Because space is frequently contiguously used and
freed, with contiguous-allocation allocation, extents,
or clustering
Keep address of first free block and count of
following free blocks
Free space list then has entries containing
addresses and counts
82

 Space Maps
Used in ZFS
Consider meta-data I/O on very large file systems
Full data structures like bit maps couldn’t fit in memory ->
thousands of I/Os
Divides device space into metaslab units and manages metaslabs
Given volume can contain hundreds of metaslabs
Each metaslab has associated space map
Uses counting algorithm
But records to log file rather than file system
Log of all block activity, in time order, in counting format
Metaslab activity -> load space map into memory in balanced-tree
structure, indexed by offset
Replay log into that structure
Combine contiguous free blocks into single entry 83

Efficiency and Performance
Efficiency dependent on:
Disk allocation and directory algorithms
Types of data kept in file’s directory
entry
Pre-allocation or as-needed allocation of
metadata structures
Fixed-size or varying-size data structures
84

Efficiency and Performance
 Performance
Keeping data and metadata close together
Buffer cache – separate section of main memory for frequently
used blocks
Synchronous writes sometimes requested by apps or needed
by OS
No buffering / caching – writes must hit disk before
acknowledgement
Asynchronous writes more common, buffer-able, faster
Free-behind and read-ahead – techniques to optimize
sequential access
Reads frequently slower than writes
85

Page Cache
A page cache caches
pages rather than disk
blocks using virtual
memory techniques and
addresses
Memory-mapped I/O
uses a page cache
Routine I/O through the
file system uses the
buffer (disk) cache
This leads to the
following figure
I/O Without a UnifiedBuffer Cache
86

Unified Buffer Cache
A unified buffer cache
uses the same page
cache to cache both
memory-mapped pages
and ordinary file system
I/O to avoid double
caching
 But which caches get
priority, and what
replacement algorithms
to use? I/O Using a Unified Buffer Cache
87

Recovery
Consistency checking – compares data in directory
structure with data blocks on disk, and tries to fix
inconsistencies
Can be slow and sometimes fails
Use system programs to back up data from disk to another
storage device (magnetic tape, other magnetic disk,
optical)
Recover lost file or disk by restoring data from backup
88

Log Structured File Systems
 Log structured (or journaling) file systems record each metadata
update to the file system as a transaction
 All transactions are written to a log
 A transaction is considered committed once it is written to the
log (sequentially)
Sometimes to a separate device or section of disk
However, the file system may not yet be updated
 The transactions in the log are asynchronously written to the file
system structures
 When the file system structures are modified, the transaction is
removed from the log
 If the file system crashes, all remaining transactions in the log must
still be performed
 Faster recovery from crash, removes chance of inconsistency of
metadata 89

The Sun Network File System (NFS)
An implementation and a specification of a software system for
accessing remote files across LANs (or WANs)
The implementation is part of the Solaris and SunOS operating
systems running on Sun workstations using an unreliable
datagram protocol (UDP/IP protocol and Ethernet
Interconnected workstations viewed as a set of independent
machines with independent file systems, which allows sharing
among these file systems in a transparent manner
A remote directory is mounted over a local file system
directory
The mounted directory looks like an integral subtree of
the local file system, replacing the subtree descending
from the local directory
90

NFS
Specification of the remote directory for the mount
operation is nontransparent; the host name of the remote
directory has to be provided
Files in the remote directory can then be accessed in a
transparent manner
Subject to access-rights accreditation, potentially any file
system (or directory within a file system), can be mounted
remotely on top of any local directory
Three Independent
File Systems 91

NFS
 NFS is designed to operate in a
heterogeneous environment of different
machines, operating systems, and network
architectures; the NFS specifications
independent of these media
 This independence is achieved through the
use of RPC primitives built on top of an
External Data Representation (XDR) protocol
used between two implementation-
independent interfaces
 The NFS specification distinguishes between
the services provided by a mount
mechanism and the actual remote-file-
access services
Mounting in NFS
92

NFS Mount Protocol
 Establishes initial logical connection between server and client
 Mount operation includes name of remote directory to be mounted
and name of server machine storing it
Mount request is mapped to corresponding RPC and forwarded to
mount server running on server machine
Export list – specifies local file systems that server exports for
mounting, along with names of machines that are permitted to
mount them
 Following a mount request that conforms to its export list, the server
returns a file handle—a key for further accesses
 File handle – a file-system identifier, and an inode number to identify
the mounted directory within the exported file system
 The mount operation changes only the user’s view and does not
affect the server side
93

NFS Protocol
Provides a set of remote procedure calls for remote file
operations. The procedures support the following operations:
searching for a file within a directory
reading a set of directory entries
manipulating links and directories
accessing file attributes
reading and writing files
NFS servers are stateless; each request has to provide a full set
of arguments (NFS V4 is just coming available – very different,
stateful)
Modified data must be committed to the server’s disk before
results are returned to the client (lose advantages of caching)
The NFS protocol does not provide concurrency-control
mechanisms
94

Three Major Layers of NFS Architecture
UNIX file-system interface (based on the open,
read, write, and close calls, and file descriptors)
Virtual File System (VFS) layer – distinguishes local
files from remote ones, and local files are further
distinguished according to their file-system types
The VFS activates file-system-specific operations to
handle local requests according to their file-system
types
Calls the NFS protocol procedures for remote requests
NFS service layer – bottom layer of the architecture
Implements the NFS protocol
95

Schematic View of NFS Architecture
96

NFS Path-Name Translation
Performed by breaking the path into component names
and performing a separate NFS lookup call for every pair of
component name and directory vnode
To make lookup faster, a directory name lookup cache on
the client’s side holds the vnodes for remote directory
names
97

NFS Remote Operations
Nearly one-to-one correspondence between regular UNIX
system calls and the NFS protocol RPCs (except opening and
closing files)
NFS adheres to the remote-service paradigm, but employs
buffering and caching techniques for the sake of performance
File-blocks cache – when a file is opened, the kernel checks with
the remote server whether to fetch or revalidate the cached
attributes
Cached file blocks are used only if the corresponding cached
attributes are up to date
File-attribute cache – the attribute cache is updated whenever
new attributes arrive from the server
Clients do not free delayed-write blocks until the server
confirms that the data have been written to disk 98

Example: WAFL File System
Used on Network Appliance “Filers” – distributed file
system appliances
“Write-anywhere file layout”
Serves up NFS, CIFS, http, ftp
Random I/O optimized, write optimized
NVRAM for write caching
Similar to Berkeley Fast File System,
with extensive modifications
99

I/O System Overview
 I/O management is a major component of operating system design
and operation
Important aspect of computer operation
I/O devices vary greatly
Various methods to control them
Performance management
New types of devices frequent
 I/ O device technology exhibits 2 conflicting
Increase standardization of software and hardware interface.
Increase broad variety of I/O devices.
 Basic I/O :Ports, busses, device controllers connect to various devices
 Device drivers encapsulate device details
Present uniform device-access interface to I/O subsystem
System call provide interface between the application and OS.
100

I/O Hardware
 Incredible variety of I/O devices
 Storage (disk, tape)
 Transmission(network connection, Bluetooth)
 Human-interface (Keyboard , mouse, Screen and Audio)
 Common concepts – signals from I/O devices interface with computer
 Port – connection point for device
 Bus – device share a common set of wires (defined a protocol – set message that can
sent on wire)
 Bus - daisy chain or shared direct access
PCI bus common in PCs and servers, PCI Express(PCIe) (connect the processorand
M/y subsystem to fast device) (16GB throughput) per second)
expansion bus connects relatively slow devices (keyboard and USB port)
 Controller (host adapter) – electronics that operate port, bus, device
Sometimes integrated
Sometimes separate circuit board (host adapter)
Contains processor,microcode, private memory, bus controller, etc
Some talk to per-device controller with bus controller,microcode,memoryetc
101

A Typical PC Bus Structure
102

I/O Hardware
I/O instructions control devices
Devices usually have registers where device driver places
commands, addresses, and data to write, or read data from
registers after command execution
Data-in register, data-out register, status register, control
register
Typically 1-4 bytes, or FIFO buffer
Devices have addresses, used by
Direct I/O instructions
Memory-mapped I/O
Device data and command registers mapped to processor
address space
Especially for large address spaces (graphics) 103

Device I/O Port Locations on PCs
104

Polling
 For each byte of I/O
 Read busy bit from status register until 0
 Host sets read or write bit and if write copies data into data-out register
 Host sets command-ready bit
 Controller sets busy bit, executes transfer
 Controller clears busy bit, error bit, command-ready bit when transfer done
 Step 1 is busy-wait cycle to wait for I/O from device
Reasonable if device is fast
But inefficient if device slow
CPU switches to other tasks?
But if miss a cycle data overwritten / lost
H/W controller to notify device become is ready or require poll
Interrupt - H/W Mechanism that enables a device to notify the CPU
105

Interrupts
 Polling can happen in 3 instruction cycles
Read status, logical-and to extract status bit, branch if not zero
How to be more efficient if non-zero infrequently?
 CPU Interrupt-request line triggered by I/O device
Checked by processor after each instruction
 Interrupt handler receives interrupts
Maskable to ignore or delay some interrupts
 Interrupt vector to dispatch interrupt to correct handler (it reduce need
of single interrupt vector – to determine which one need service)
Context switch at start and end
Based on priority
Some nonmaskable
Interrupt chaining if more than one device at same interrupt number
106

Interrupt-Driven I/O Cycle
Interrupt Priority: Low priority without
masking
High priority preempt execution of
low priority
107

Intel Pentium Processor Event-Vector Table
108

Interrupts
Interrupt mechanism also used for exceptions
Terminate process, crash system due to hardware error
Page fault executes when memory access error
System call executes via trap (software interrupt) to
trigger kernel to execute request
Multi-CPU systems can process interrupts
concurrently
If operating system designed to handle it
Used for time-sensitive processing, frequent, must
be fast
109

Direct Memory Access
 Used to avoid programmedI/O (one byte at a time) for large data movement
 Many computer avoid burdening main CPU by PIO by offloading of this work for
special processor called DMA
 Requires DMA controller
 Bypasses CPU to transfer data directly between I/O device and memory
 OS writes DMA command block into memory
 Source and destination addresses
 Read or write mode
 Count of bytes
 Writes location of commandblock to DMA controller
 Bus mastering of DMA controller – grabs bus from CPU
Cycle stealing from CPU but still much more efficient
 When done, interrupts to signal completion
 Version that is aware of virtual addresses can be even more efficient - DVMA
110

Six Step Process to Perform DMA Transfer
111

Application I/O Interface
How app can open file on the disk without knowing what
kind of disk
How new disk added to computer without disturbing the
OS
Interface - I/O system calls encapsulate device behaviors in
generic classes
Device-driver layer hides differences among I/O controllers
from kernel
New devices talking already-implemented protocols need
no extra work
Each OS has its own I/O subsystem structures and device
driver frameworks
112

 Devices vary in many dimensions
Character-stream or block (transfer bytes one by one , block of
bytes as a unit)
Sequential or random-access (transfer data in a fixed order
determined, instruct the device to seek of the available data
storage location)
Synchronous or asynchronous (or both) ( transfer with predictate
response time,(or) irregular, unpredictable )
Sharable or dedicated ( can be used concurrently by several
process)
Speed of operation ( range from a few bytes per/sec to few GB )
read-write, read only, or write only( perform both input and
output
 but other support only one data transfer direction)
113

Characteristics of I/O Devices
115

Characteristics of I/O Devices
 Subtleties of devices handled by device drivers
 Broadly I/O devices can be grouped by the OS into
 Block I/O
 Character I/O (Stream)
 Memory-mappedfile access
 Networksockets
 For direct manipulation of I/O device specificcharacteristics,usually an escape /
back door
 Unix ioctl() call to send arbitrary bits to a device control register and data
to device data register
 3 systemcall
 File descript (connect app by referring hardware device)
 Integer (select one of the command)
 Pointer (arbitrary DS in m/y enable necessary application and driver to
communicate any necessary control informationdata)
116

Block and Character Devices
 Block devices include disk drives
Commands include read, write, seek
Raw I/O (access the block device as simple linear array of block)
direct I/O, or file-system access ( OS allow a mode of operation on
a file disable buffering and locking)
Memory-mapped file access possible
File mapped to virtual memory and clusters brought via
demand paging
DMA
 Character devices include keyboards, mice, serial ports
Commands include get(), put()
Libraries layered on top allow line editing
117

Network Devices
Varying enough from block and character to have own
interface
Linux, Unix, Windows and many others include socket
interface
Separates network protocol from network operation
Includes select() functionality
Approaches vary widely (pipes, FIFOs, streams, queues,
mailboxes)
118

Clocks and Timers
Provide current time, elapsed time, timer
Normal resolution about 1/60 second
Some systems provide higher-resolutiontimers
Programmable interval timer used for timings,
periodic interrupts
ioctl() (on UNIX) covers odd aspects of I/O such as
clocks and timers
119

Nonblocking and Asynchronous I/O
 Blocking - process suspended until I/O completed
 Easy to use and understand
 Insufficient for some needs
 Nonblocking - I/O call returns as much as available
 User interface,data copy (buffered I/O)
 Implemented via multi-threading
 Returns quickly with count of bytes read or written
 select() to find if data ready then read() or write() to transfer eg.
Select();(application poll for n/w activity without blocking)
 Asynchronous - process runs while I/O executes
 Difficultto use
 I/O subsystemsignals process when I/O completed
 Eg. Read() cal request transfer will performedin its entirety but will
complete at some future time
120

Two I/O Methods
Synchronous Asynchronous
121

Vectored I/O
Vectored I/O allows one system call to perform
multiple I/O operations involve multiple location
For example, Unix readve() accepts a vector of
multiple buffers to read into or write from
This scatter-gather method better than multiple
individual I/O calls
Decreases context switching and system call overhead
Some versions provide atomicity
Avoid for example worry about multiple threads changing
data as reads / writes occurring
122

Kernel I/O Subsystem
 I/O Scheduling
 Some I/O request ordering via per-device queue
 Some OSs try fairness
 Some implementQuality Of Service (i.e. IPQOS)
 Device Status Table - Kernel support asynchronus I/O , must be able to keep
track of many I/O request at same time.
 Table indicate – device type, address and state ,type of request.
 Buffering - store data in memory while transferring betweendevices
 To cope with device speed mismatch (producer and consumer)
 To cope with device transfer size mismatch
 To maintain “copy semantics”
 Double buffering – two copies of the data (2 buffer, 1-filled, 2 write)
Kernel and user , Varying sizes , Full / being processed and not-full /
being used , Copy-on-write can be used for efficiency in some cases
123

Kernel I/O Subsystem
 Buffer - it hold only existing copy of a data item
 Caching - faster device holding copy of data
Always just a copy (by definition, holds the only existing copy of
data item that resides elsewhere.
Buffer also used as a cache - Key to I/O performance
Sometimes combined with buffering
 Spooling – is a buffer that hold output for a device
If device can serve only one request at a time
i.e., Printing
Spooling system copies the queued spool files for output to the
printer.
 Device reservation - provides exclusive access to a device
System calls for allocation and de-allocation
Watch out for deadlock 125

Error Handling
 OS uses protected m/y against many kind oh h/w and application
errors.
 If minor malfunction – computer system failure is not usual result.
 Device and I/o can fail in many ways
 Transient reason - n/w become overheaded
 Permanent reason – disk controller become defective
 OS can recover from disk read, device unavailable, transient write
failures
 Retry a read or write, send (retry / resend), for example
 Some systems more advanced – Solaris FMA, AIX
 Track error frequencies, stop using device with increasing
frequency of retry-able errors
126

General rule , - one bit information about status
call (success/ failure)
Errno (system call) – used to return error code –
indicate failure ( eg. Argument out, of range , bad
pointer , file not open)
Failure in SCSI device reported 3 level
Sense key – identify general natural failure(h/w error or
illegal request)
Additional sense code – state the category of failure
(bad command, self – test failure)
Additional sense code qualifier – even give more detail (
which command parameter was in error or which h/w
subsystem failure) 127

I/O Protection
Errors are closely related to issue of protection.
User process may accidentally or purposefully
attempt to disrupt normal operation via illegal I/O
instructions
To prevent user from performing illegal I/O
All I/O instructions defined to be privileged
User can’t directly use i/O through only OS
I/O must be performed via system calls
Memory-mapped and I/O port memory
locations must be protected too
128

Use of a System Call to Perform I/O
129

Kernel Data Structures
 Kernel keeps state info for I/O components, including open file tables,
network connections, character device state (open file table structure)
 Many, many complex data structures to track buffers, memory allocation,
“dirty” blocks
 UNIX – variety of entities , such as user file, raw file and address space of
process
 Kernel need buffer cache before performing I/O (it ensure the request size
based on sector boundary)
 Some use object-oriented methods and message passing to implement I/O
Windows uses message passing
Message with I/O information passed from user mode into kernel
Message modified as it flows through to device driver and back to
process
For Output – msg contain the data to be written
For Input - msg contain buffer to receive the data
130

Power Management
Not strictly domain of I/O, but much is I/O related
Computers and devices use electricity, generate
heat, frequently require cooling
OSes can help manage and improve use
Cloud computing environments move virtual
machines between servers
Can end up evacuating whole systems and
shutting them down
Mobile computing has power management as first
class OS aspect
132

Power Management
 For example, Android implements
Component-level power management
Understands relationship between components
Build device tree representing physical device topology
System bus -> I/O subsystem -> {flash, USB storage}
Device driver tracks state of device, whether in use
Unused component – turn it off
All devices in tree branch unused – turn off branch
Wake locks – like other locks but prevent sleep of device when
lock is held
Power collapse – put a device into very deep sleep
Marginal power use
Only awake enough to respond to external stimuli (button
press, incoming call) 133

I/O Requests to Hardware Operations
 Consider reading a file from disk
for a process:
Determine device holding file
Translate name to device
representation
Physically read data from disk
into buffer
Make data available to
requesting process
Return control to process
Life Cycle of An I/O Request
134

STREAMS
 STREAM – a full-duplex communication channel between a user-level
process and a device in Unix System V and beyond
 A STREAM consists of:
STREAM head interfaces with the user process
driver end interfaces with the device
zero or more STREAM modules between them
 Each module contains a read queue and a write queue
 Message passing is used to communicate between queues
Flow control option to indicate available or busy
 Asynchronous internally, synchronous where user process
communicates with stream head
135

Performance
I/O a major factor in
system performance:
Demands CPU to
execute device
driver, kernel I/O
code
Context switches
due to interrupts
Data copying
Network traffic
especially stressful
Intercomputer Communications
137

Improving Performance
Reduce number of context switches
Reduce data copying
Reduce interrupts by using large transfers, smart
controllers, polling
Use DMA
Use smarter hardware devices
Balance CPU, memory, bus, and I/O performance
for highest throughput
Move user-mode processes / daemons to kernel
threads
138

Device-Functionality Progression
139

Cs8493 unit 4

More Related Content

What's hot (20)

Similar to Cs8493 unit 4 (20)

More from Kathirvel Ayyaswamy (20)

Recently uploaded (20)

Cs8493 unit 4