1 ==============================
2 Deadline IO scheduler tunables
3 ==============================
4
5 This little file attempts to document how the
6 In particular, it will clarify the meaning of
7 of interest to power users.
8
9 Selecting IO schedulers
10 -----------------------
11 Refer to Documentation/block/switching-sched.r
12 selecting an io scheduler on a per-device basi
13
14 ----------------------------------------------
15
16 read_expire (in ms)
17 -----------------------
18
19 The goal of the deadline io scheduler is to at
20 service time for a request. As we focus mainly
21 tunable. When a read request first enters the
22 a deadline that is the current time + the read
23 milliseconds.
24
25
26 write_expire (in ms)
27 -----------------------
28
29 Similar to read_expire mentioned above, but fo
30
31
32 fifo_batch (number of requests)
33 ------------------------------------
34
35 Requests are grouped into ``batches`` of a par
36 write) which are serviced in increasing sector
37 deadline expiries are only checked between bat
38 maximum number of requests per batch.
39
40 This parameter tunes the balance between per-r
41 throughput. When low latency is the primary c
42 a value of 1 yields first-come first-served be
43 generally improves throughput, at the cost of
44
45
46 writes_starved (number of dispatches)
47 --------------------------------------
48
49 When we have to move requests from the io sche
50 device dispatch queue, we always give a prefer
51 don't want to starve writes indefinitely eithe
52 how many times we give preference to reads ove
53 done writes_starved number of times, we dispat
54 same criteria as reads.
55
56
57 front_merges (bool)
58 ----------------------
59
60 Sometimes it happens that a request enters the
61 with a request that is already on the queue. E
62 request, or it fits at the front. That is call
63 or a front merge candidate. Due to the way fil
64 back merges are much more common than front me
65 may even know that it is a waste of time to sp
66 front merge requests. Setting front_merges to
67 Front merges may still occur due to the cached
68 that comes at basically 0 cost we leave that o
69 rbtree front sector lookup when the io schedul
70
71
72 Nov 11 2002, Jens Axboe <jens.axboe@oracle.com>
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.