Operating Systems

ECE344
Lecture 11: File System

Ding Yuan
Big picture: performance
gap

ECE344 - Operating Systems - Ding

2
Yuan
 What problem are we
solving?
• Data storage & access
• Super important

• One of the fastest

growing industry
• Why? 1956, IBM, 24 inches, 3.75MB,
• Driven by technology 1KB/sec, > $150,000

2015, Seagate, 3.5

inches, 5TB, 6Gb/sec, <
$200

ECE344 - Operating Systems - Ding

3
Yuan
 What problem are we
solving?
• One of the fastest growing industry
• Why?
• Driven by technology
• Driven by demand
• Mainframe storage: IBM, Memorex
• PC storage: Seagate, DEC, Quantum, etc.
• Enterprise Storage: EMC, NetApp, etc.
• Cloud Storage: Dropbox, Google Drive, etc.

ECE344 - Operating Systems - Ding

4
Yuan
 History of storage tech

1956: IBM 350, 24 inches, 1980: IBM 3380, first GB 1980: ST-560, first 5.25
3.75MB, 1KB/sec, > $150,000
disk (1.26G), > $100K inch drive, 5MB, $1500

NextCom notebook 2006: first

Tape (DECtape): primary storage for laptop w/ SSD as storage
main-frames and mini-computers (19505 ~ 70s)
ECE344 - Operating Systems - Ding
Yuan
 6
Image source: https://en.wikipedia.org/wiki/File:SixHardDriveFormFactors.jpg/
 Storage Industry: The origin
of “disruptive technology”

Image source: https://en.wikipedia.org/wiki/Disruptive_innovation

Image source: amazon.ca

7
 File Systems
• First we’ll discuss properties of physical disks
• Structure
• Performance
• Scheduling

• Then we’ll discuss how we build file systems on them

• Files
• Directories
• Sharing
• Protection
• File System Layouts
• File Buffer Cache
• Read Ahead

• Finally, we will discuss SSD

ECE344 - Operating Systems - Ding
8
Yuan
 Disks and the OS
• Disks are messy physical devices
• Errors, bad blocks, missed seeks, etc.

• The job of the OS is to hide this mess from

higher level software
• Low-level device control (initiate a disk read,
etc.)
• Higher-level abstractions (files, databases,
etc.)

ECE344 - Operating Systems - Ding

9
Yuan
 How hard disk work?
• http://www.youtube.com/watch?v=kdmLvl1n82
U
• https://youtu.be/9eMWG3fwiEU?t=11

• Disk components
• Platters
• Surfaces
• Tracks
• Cylinders
• Sectors
• Arm
• Heads
ECE344 - Operating Systems - Ding
10
Yuan
Another View of Disk

ECE344 - Operating Systems - Ding

11
Yuan
 Disk Interaction
• Specifying disk requests requires a lot of info:
• Cylinder #, surface #, sector #, transfer size…

• Older disks required the OS to specify all of this

• The OS needed to know all disk parameters

• Modern disks are more complicated

• Not all sectors are the same size, sectors are remapped, etc.

• Current disks provide a higher-level interface (SCSI)

• The disk exports its data as a logical array of blocks [0…N]
• Disk maps logical blocks to cylinder/surface/track/sector
• Only need to specify the logical block # to read/write
• But now the disk parameters are hidden from the OS
ECE344 - Operating Systems - Ding
12
Yuan
 Disk Performance
• Random disk access is SLOW!
Access data
Data sequentially: only
suffer one seek and
rotational delay

Rotational
delay
Seek Random disk access:
suffers one seek and
rotational delay
every time!

ECE344 - Operating Systems - Ding

13
Yuan
 Disk Performance
• Disk request performance depends upon three steps
• Seek – moving the disk arm to the correct cylinder
• Depends on how fast disk arm can move (increasing very
slowly)
• Rotation – waiting for the sector to rotate under the head
• Depends on rotation rate of disk (increasing, but slowly)
• Transfer – transferring data from surface into disk
controller electronics, sending it back to the host
• Depends on density (increasing quickly)

• When the OS uses the disk, it tries to minimize the

cost of all of these steps
• Particularly seeks and rotation
ECE344 - Operating Systems - Ding
14
Yuan
 Disks: 2016
• Seagate Cheetah 3.5" (server)
• capacity: 300 - 600 GB
• rotational speed: 15,000 RPM
• sequential read performance: 122 MB/s - 204
MB/s
• seek time (average): 3.4 ms

• Seagate Barracuda 3.5" (desktop)

• capacity: 250 GB – 4TB
• rotational speed: 7,200 RPM
• sequential read performance: 125 MB/s - 146
MB/s 15
ECE344 - Operating Systems - Ding
Yuan
 Disk Scheduling
• Because seeks are so expensive (milliseconds!), the
OS tries to schedule disk requests that are queued
waiting for the disk
• FCFS (do nothing)
• Reasonable when load is low
• Long waiting times for long request queues
• SSTF (shortest seek time first)
• Minimize arm movement (seek time), maximize request rate
• Favors middle blocks
• SCAN (elevator)
• Service requests in one direction until done, then reverse
• C-SCAN
• Like SCAN, but only go in one direction (typewriter)
ECE344 - Operating Systems - Ding
16
Yuan
 ECE344 - Operating Systems - Ding
17
Yuan
 Disk Scheduling (2)

• In general, unless there are large request

queues, disk scheduling does not have much
impact
• Important for servers, less so for PCs

• Modern disks often do the disk scheduling

themselves
• Disks know their layout better than OS, can
optimize better
• Ignores, undoes any scheduling done by OS

ECE344 - Operating Systems - Ding

18
Yuan
Stages of I/O Request

ECE344 - Operating Systems - Ding

19
Yuan
 But do you directly
program on “disk”?
Life with an OS Life without an OS
file = open (“test.txt”, • Where is this file on disk?
O_WRONLY); Which platter, track, and
write (file, “test”, 4); sectors?

close (file); • Code needs to change on a

different system

ECE344 - Operating Systems - Ding

20
Yuan
 File Systems

• File systems
• Implement an abstraction (files) for secondary
storage
• Organize files logically (directories)
• Permit sharing of data between processes,
people, and machines
• Protect data from unwanted access (security)

ECE344 - Operating Systems - Ding

21
Yuan
 Files
• A file is data with some properties
• Contents, size, owner, last read/write time, protection, etc.

• A file can also have a type

• Understood by other parts of the OS or runtime libraries
• Executable, dll, souce, object, text, etc.
• Understood by the file system
• Block/character device, directory, link, etc.

• A file’s type can be encoded in its name or contents

• Windows encodes type in name
• .com, .exe, .bat, .dll, .jpg, etc.
• Unix encodes type in contents
• Magic numbers, initial characters (e.g., #! for shell scripts)
ECE344 - Operating Systems - Ding
22
Yuan
 Basic File Operations
Unix Windows

• • CreateFile(name, CREATE)
creat(name)
• CreateFile(name, OPEN)
• open(name, how)
• ReadFile(handle, …)
• read(fd, buf, len)
• WriteFile(handle, …)
• write(fd, buf, len)
• FlushFileBuffers(handle, …)
• sync(fd) • SetFilePointer(handle, …)

• seek(fd, pos) • CloseHandle(handle, …)

• close(fd) • DeleteFile(name)

• CopyFile(name)
• unlink(name)
• MoveFile(name)
ECE344 - Operating Systems - Ding
23
Yuan
 Directories
• Directories serve two purposes
• For users, they provide a structured way to organize files
• For the file system, they provide a convenient naming
interface that allows the implementation to separate logical
file organization from physical file placement on the disk

• Most file systems support multi-level directories

• Naming hierarchies (/, /usr, /usr/local/, …)

• Most file systems support the notion of a current

directory
• Relative names specified with respect to current directory
• Absolute names start from the root of directory tree

ECE344 - Operating Systems - Ding

24
Yuan
 Directory Internals
• A directory is a list of entries
• <name, location>
• Name is just the name of the file or directory
• Location depends upon how file is represented
on disk

• List is usually unordered (effectively random)

• Entries usually sorted by program that reads
directory

• Directories typically stored in files

ECE344 - Operating Systems - Ding

25
Yuan
 Basic Directory
Operations
Unix NT

• Directories implemented in • Explicit dir operations

files • CreateDirectory(name)
• Use file ops to create dirs • RemoveDirectory(name
)
• C runtime library provides a
higher-level abstraction for • Very different method
reading directories for reading directory
• opendir(name) entries
• readdir(DIR) • FindFirstFile(pattern)
• seekdir(DIR) • FindNextFile()
• closedir(DIR)

ECE344 - Operating Systems - Ding

26
Yuan
 Path Name Translation
• Let’s say you want to open “/one/two/three”

• What does the file system do?

• Open directory “/” (well known, can always find)
• Search for the entry “one”, get location of “one” (in dir entry)
• Open directory “one”, search for “two”, get location of “two”
• Open directory “two”, search for “three”, get location of
“three”
• Open file “three”

• Systems spend a lot of time walking directory paths

• This is why open is separate from read/write
• OS will cache prefix lookups for performance
• /a/b, /a/bb, /a/bbb, etc., all share “/a” prefix
ECE344 - Operating Systems - Ding
27
Yuan
 File System Layout
How do file systems use the disk to store files?
• File systems define a block size (e.g., 4KB)
• Disk space is allocated in granularity of blocks

• A “Master Block” determines location of root directory

• Always at a well-known disk location
• Often replicated across disk for reliability

• A free map determines which blocks are free, allocated

• Usually a bitmap, one bit per block on the disk
• Also stored on disk, cached in memory for performance

• Remaining disk blocks used to store files (and dirs)

• There are many ways to do this
ECE344 - Operating Systems - Ding
28
Yuan
 Disk Layout Strategies
• Files span multiple disk blocks

• How do you find all of the blocks for a file?

1. Contiguous allocation
• Fast, simplifies directory access
• Inflexible, causes fragmentation, needs compaction
2. Linked structure
• Each block points to the next, directory points to the first
• Good for sequential access, bad for all others
3. Indexed structure (indirection, hierarchy)
• An “index block” contains pointers to many other blocks
• Handles random better, still good for sequential
• May need multiple index blocks (linked together)
ECE344 - Operating Systems - Ding
29
Yuan
Logical layout: inode (index
node)

30
 Unix Inodes
• Unix inodes implement an indexed structure for files
• Also store metadata info (protection, timestamps, length, ref count…)

• Each inode contains 15 block pointers

• First 12 are direct blocks (e.g., 4 KB blocks)
• Then single, double, and triple indirect

(Metadata)
0
1
…
…
1 (1)
1
2 (2)
1
3 (3)
4 … …
ECE344 - Operating Systems - Ding
31
Yuan
 Unix Inodes and Path
Search
• Unix Inodes are not directories

• Inodes describe where on the disk the blocks for a file are
placed
• Directories are files, so inodes also describe where the blocks
for directories are placed on the disk

• Directory entries map file names to inodes

• To open “/one”, use Master Block to find inode for “/” on disk
• Open “/”, look for entry for “one”
• This entry gives the inode number for the inode for “one”
• Read the inode for “one” into memory
• The inode says where first data block is on disk
• Read that block into memory to access the data in the file
ECE344 - Operating Systems - Ding
32
Yuan
 Example: read /bin/ls
• /bin/ls

inode inode of /bin

of / usr 2783 inode of /bin/ls
bin 3512 cd 5193
… .. .. ls 5728
.. .. instructions
of ls.. ..

instructions
of ls.. ..

ECE344 - Operating Systems - Ding

33
Yuan
 Sharing Files btw.
Directories
• Links (or hard links)
• ln source_file target_dir
• Simply create another link from target_dir to the
inode of source_file (the inode is not duplicated)
• Now two directories have links to source_file
• What if we remove one?
• Now you understand why the system call to
remove a file is named “unlink”?

ECE344 - Operating Systems - Ding

34
Yuan
 File Buffer Cache
• Applications exhibit significant locality for reading and
writing files
• Idea: Cache file blocks in memory to capture locality
• This is called the file buffer cache
• Cache is system wide, used and shared by all processes
• Reading from the cache makes a disk perform like memory
• Even a 4 MB cache can be very effective

• Issues
• The file buffer cache competes with VM (tradeoff here)
• Like VM, it has limited size
• Need replacement algorithms again (LRU usually used)

ECE344 - Operating Systems - Ding

35
Yuan
 Caching Writes
• On a write, some applications assume that data
makes it through the buffer cache and onto the
disk
• As a result, writes are often slow even with caching

• Several ways to compensate for this

• “write-behind”
• Maintain a queue of uncommitted blocks
• Periodically flush the queue to disk
• Unreliable
• Battery backed-up RAM (NVRAM)
• As with write-behind, but maintain queue in NVRAM
• Expensive
ECE344 - Operating Systems - Ding
36
Yuan
 Performance Issues
Original Unix FS had two placement problems:
1. Data blocks allocated randomly in aging file
systems

Blocks for the same file allocated sequentially when
FS is new

As FS “ages” and fills, need to allocate into blocks
freed up when other files are deleted

Problem: Deleted files essentially randomly placed

So, blocks for new files become scattered across the
disk
2. Inodes allocated far from blocks

All inodes at beginning of disk, far from data

Traversing file name paths, manipulating files,
directories requires going back and forth from inodes
ECE344 - Operating Systems - Ding
to data blocks 37
Yuan
FreeBSD Fast File
System

38
 Fast File System
• BSD FFS addressed these problems using the notion
of a cylinder group
• Disk partitioned into groups of cylinders
• Data blocks in same file allocated in same cylinder
• Files in same directory allocated in same cylinder
• Inodes for files allocated in same cylinder as file data
blocks

• Free space requirement

• To be able to allocate according to cylinder groups, the
disk must have free space scattered across cylinders
• 10% of the disk is reserved just for this purpose

ECE344 - Operating Systems - Ding

39
Yuan