Based on kernel version 2.6.26. Page generated on 2008-07-16 21:13 EST.
1 I/O statistics fields 2 --------------- 3 4 Last modified Sep 30, 2003 5 6 Since 2.4.20 (and some versions before, with patches), and 2.5.45, 7 more extensive disk statistics have been introduced to help measure disk 8 activity. Tools such as sar and iostat typically interpret these and do 9 the work for you, but in case you are interested in creating your own 10 tools, the fields are explained here. 11 12 In 2.4 now, the information is found as additional fields in 13 /proc/partitions. In 2.6, the same information is found in two 14 places: one is in the file /proc/diskstats, and the other is within 15 the sysfs file system, which must be mounted in order to obtain 16 the information. Throughout this document we'll assume that sysfs 17 is mounted on /sys, although of course it may be mounted anywhere. 18 Both /proc/diskstats and sysfs use the same source for the information 19 and so should not differ. 20 21 Here are examples of these different formats: 22 23 2.4: 24 3 0 39082680 hda 446216 784926 9550688 4382310 424847 312726 5922052 19310380 0 3376340 23705160 25 3 1 9221278 hda1 35486 0 35496 38030 0 0 0 0 0 38030 38030 26 27 28 2.6 sysfs: 29 446216 784926 9550688 4382310 424847 312726 5922052 19310380 0 3376340 23705160 30 35486 38030 38030 38030 31 32 2.6 diskstats: 33 3 0 hda 446216 784926 9550688 4382310 424847 312726 5922052 19310380 0 3376340 23705160 34 3 1 hda1 35486 38030 38030 38030 35 36 On 2.4 you might execute "grep 'hda ' /proc/partitions". On 2.6, you have 37 a choice of "cat /sys/block/hda/stat" or "grep 'hda ' /proc/diskstats". 38 The advantage of one over the other is that the sysfs choice works well 39 if you are watching a known, small set of disks. /proc/diskstats may 40 be a better choice if you are watching a large number of disks because 41 you'll avoid the overhead of 50, 100, or 500 or more opens/closes with 42 each snapshot of your disk statistics. 43 44 In 2.4, the statistics fields are those after the device name. In 45 the above example, the first field of statistics would be 446216. 46 By contrast, in 2.6 if you look at /sys/block/hda/stat, you'll 47 find just the eleven fields, beginning with 446216. If you look at 48 /proc/diskstats, the eleven fields will be preceded by the major and 49 minor device numbers, and device name. Each of these formats provide 50 eleven fields of statistics, each meaning exactly the same things. 51 All fields except field 9 are cumulative since boot. Field 9 should 52 go to zero as I/Os complete; all others only increase. Yes, these are 53 32 bit unsigned numbers, and on a very busy or long-lived system they 54 may wrap. Applications should be prepared to deal with that; unless 55 your observations are measured in large numbers of minutes or hours, 56 they should not wrap twice before you notice them. 57 58 Each set of stats only applies to the indicated device; if you want 59 system-wide stats you'll have to find all the devices and sum them all up. 60 61 Field 1 -- # of reads completed 62 This is the total number of reads completed successfully. 63 Field 2 -- # of reads merged, field 6 -- # of writes merged 64 Reads and writes which are adjacent to each other may be merged for 65 efficiency. Thus two 4K reads may become one 8K read before it is 66 ultimately handed to the disk, and so it will be counted (and queued) 67 as only one I/O. This field lets you know how often this was done. 68 Field 3 -- # of sectors read 69 This is the total number of sectors read successfully. 70 Field 4 -- # of milliseconds spent reading 71 This is the total number of milliseconds spent by all reads (as 72 measured from __make_request() to end_that_request_last()). 73 Field 5 -- # of writes completed 74 This is the total number of writes completed successfully. 75 Field 7 -- # of sectors written 76 This is the total number of sectors written successfully. 77 Field 8 -- # of milliseconds spent writing 78 This is the total number of milliseconds spent by all writes (as 79 measured from __make_request() to end_that_request_last()). 80 Field 9 -- # of I/Os currently in progress 81 The only field that should go to zero. Incremented as requests are 82 given to appropriate struct request_queue and decremented as they finish. 83 Field 10 -- # of milliseconds spent doing I/Os 84 This field is increases so long as field 9 is nonzero. 85 Field 11 -- weighted # of milliseconds spent doing I/Os 86 This field is incremented at each I/O start, I/O completion, I/O 87 merge, or read of these stats by the number of I/Os in progress 88 (field 9) times the number of milliseconds spent doing I/O since the 89 last update of this field. This can provide an easy measure of both 90 I/O completion time and the backlog that may be accumulating. 91 92 93 To avoid introducing performance bottlenecks, no locks are held while 94 modifying these counters. This implies that minor inaccuracies may be 95 introduced when changes collide, so (for instance) adding up all the 96 read I/Os issued per partition should equal those made to the disks ... 97 but due to the lack of locking it may only be very close. 98 99 In 2.6, there are counters for each cpu, which made the lack of locking 100 almost a non-issue. When the statistics are read, the per-cpu counters 101 are summed (possibly overflowing the unsigned 32-bit variable they are 102 summed to) and the result given to the user. There is no convenient 103 user interface for accessing the per-cpu counters themselves. 104 105 Disks vs Partitions 106 ------------------- 107 108 There were significant changes between 2.4 and 2.6 in the I/O subsystem. 109 As a result, some statistic information disappeared. The translation from 110 a disk address relative to a partition to the disk address relative to 111 the host disk happens much earlier. All merges and timings now happen 112 at the disk level rather than at both the disk and partition level as 113 in 2.4. Consequently, you'll see a different statistics output on 2.6 for 114 partitions from that for disks. There are only *four* fields available 115 for partitions on 2.6 machines. This is reflected in the examples above. 116 117 Field 1 -- # of reads issued 118 This is the total number of reads issued to this partition. 119 Field 2 -- # of sectors read 120 This is the total number of sectors requested to be read from this 121 partition. 122 Field 3 -- # of writes issued 123 This is the total number of writes issued to this partition. 124 Field 4 -- # of sectors written 125 This is the total number of sectors requested to be written to 126 this partition. 127 128 Note that since the address is translated to a disk-relative one, and no 129 record of the partition-relative address is kept, the subsequent success 130 or failure of the read cannot be attributed to the partition. In other 131 words, the number of reads for partitions is counted slightly before time 132 of queuing for partitions, and at completion for whole disks. This is 133 a subtle distinction that is probably uninteresting for most cases. 134 135 More significant is the error induced by counting the numbers of 136 reads/writes before merges for partitions and after for disks. Since a 137 typical workload usually contains a lot of successive and adjacent requests, 138 the number of reads/writes issued can be several times higher than the 139 number of reads/writes completed. 140 141 In 2.6.25, the full statistic set is again available for partitions and 142 disk and partition statistics are consistent again. Since we still don't 143 keep record of the partition-relative address, an operation is attributed to 144 the partition which contains the first sector of the request after the 145 eventual merges. As requests can be merged across partition, this could lead 146 to some (probably insignificant) innacuracy. 147 148 Additional notes 149 ---------------- 150 151 In 2.6, sysfs is not mounted by default. If your distribution of 152 Linux hasn't added it already, here's the line you'll want to add to 153 your /etc/fstab: 154 155 none /sys sysfs defaults 0 0 156 157 158 In 2.6, all disk statistics were removed from /proc/stat. In 2.4, they 159 appear in both /proc/partitions and /proc/stat, although the ones in 160 /proc/stat take a very different format from those in /proc/partitions 161 (see proc(5), if your system has it.) 162 163 -- ricklind[AT]us.ibm[DOT]com
