![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to 2.t CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [CSRG BSD Unix] / 43BSDReno / share / doc / smm / 14.fastfs|
Return to 2.t CVS log | Up to [CSRG BSD Unix] / 43BSDReno / share / doc / smm / 14.fastfs |
1.1 root 1: .\" Copyright (c) 1986 The Regents of the University of California.
2: .\" All rights reserved.
3: .\"
4: .\" Redistribution and use in source and binary forms are permitted
5: .\" provided that the above copyright notice and this paragraph are
6: .\" duplicated in all such forms and that any documentation,
7: .\" advertising materials, and other materials related to such
8: .\" distribution and use acknowledge that the software was developed
9: .\" by the University of California, Berkeley. The name of the
10: .\" University may not be used to endorse or promote products derived
11: .\" from this software without specific prior written permission.
12: .\" THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
13: .\" IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
14: .\" WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
15: .\"
16: .\" @(#)2.t 6.2 (Berkeley) 3/7/89
17: .\"
18: .ds RH Old file system
19: .NH
20: Old File System
21: .PP
22: In the file system developed at Bell Laboratories
23: (the ``traditional'' file system),
24: each disk drive is divided into one or more
25: partitions. Each of these disk partitions may contain
26: one file system. A file system never spans multiple
27: partitions.\(dg
28: .FS
29: \(dg By ``partition'' here we refer to the subdivision of
30: physical space on a disk drive. In the traditional file
31: system, as in the new file system, file systems are really
32: located in logical disk partitions that may overlap. This
33: overlapping is made available, for example,
34: to allow programs to copy entire disk drives containing multiple
35: file systems.
36: .FE
37: A file system is described by its super-block,
38: which contains the basic parameters of the file system.
39: These include the number of data blocks in the file system,
40: a count of the maximum number of files,
41: and a pointer to the \fIfree list\fP, a linked
42: list of all the free blocks in the file system.
43: .PP
44: Within the file system are files.
45: Certain files are distinguished as directories and contain
46: pointers to files that may themselves be directories.
47: Every file has a descriptor associated with it called an
48: .I "inode".
49: An inode contains information describing ownership of the file,
50: time stamps marking last modification and access times for the file,
51: and an array of indices that point to the data blocks for the file.
52: For the purposes of this section, we assume that the first 8 blocks
53: of the file are directly referenced by values stored
54: in an inode itself*.
55: .FS
56: * The actual number may vary from system to system, but is usually in
57: the range 5-13.
58: .FE
59: An inode may also contain references to indirect blocks
60: containing further data block indices.
61: In a file system with a 512 byte block size, a singly indirect
62: block contains 128 further block addresses,
63: a doubly indirect block contains 128 addresses of further singly indirect
64: blocks,
65: and a triply indirect block contains 128 addresses of further doubly indirect
66: blocks.
67: .PP
68: A 150 megabyte traditional UNIX file system consists
69: of 4 megabytes of inodes followed by 146 megabytes of data.
70: This organization segregates the inode information from the data;
71: thus accessing a file normally incurs a long seek from the
72: file's inode to its data.
73: Files in a single directory are not typically allocated
74: consecutive slots in the 4 megabytes of inodes,
75: causing many non-consecutive blocks of inodes
76: to be accessed when executing
77: operations on the inodes of several files in a directory.
78: .PP
79: The allocation of data blocks to files is also suboptimum.
80: The traditional
81: file system never transfers more than 512 bytes per disk transaction
82: and often finds that the next sequential data block is not on the same
83: cylinder, forcing seeks between 512 byte transfers.
84: The combination of the small block size,
85: limited read-ahead in the system,
86: and many seeks severely limits file system throughput.
87: .PP
88: The first work at Berkeley on the UNIX file system attempted to improve both
89: reliability and throughput.
90: The reliability was improved by staging modifications
91: to critical file system information so that they could
92: either be completed or repaired cleanly by a program
93: after a crash [Kowalski78].
94: The file system performance was improved by a factor of more than two by
95: changing the basic block size from 512 to 1024 bytes.
96: The increase was because of two factors:
97: each disk transfer accessed twice as much data,
98: and most files could be described without need to access
99: indirect blocks since the direct blocks contained twice as much data.
100: The file system with these changes will henceforth be referred to as the
101: .I "old file system."
102: .PP
103: This performance improvement gave a strong indication that
104: increasing the block size was a good method for improving
105: throughput.
106: Although the throughput had doubled,
107: the old file system was still using only about
108: four percent of the disk bandwidth.
109: The main problem was that although the free list was initially
110: ordered for optimal access,
111: it quickly became scrambled as files were created and removed.
112: Eventually the free list became entirely random,
113: causing files to have their blocks allocated randomly over the disk.
114: This forced a seek before every block access.
115: Although old file systems provided transfer rates of up
116: to 175 kilobytes per second when they were first created,
117: this rate deteriorated to 30 kilobytes per second after a
118: few weeks of moderate use because of this
119: randomization of data block placement.
120: There was no way of restoring the performance of an old file system
121: except to dump, rebuild, and restore the file system.
122: Another possibility, as suggested by [Maruyama76],
123: would be to have a process that periodically
124: reorganized the data on the disk to restore locality.
125: .ds RH New file system
126: .sp 2
127: .ne 1i
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.