The I/O scheduler optimizes disk access. Issuing each request directly to the disk results in a bad throughput, because this strategy would cause a high amount of small requests, where in the worst case each request demands a completely different disk area. Today data transfer rates from physical disks are very high, so transfer times are very short. The dominating part of the access time is the disk latency, mainly caused by disk head movements of about 8 ms in average. The strategy for optimization aims at minimizing the number of I/O operations and disk head movements.
I/O schedulers mostly implement two functions:
*
Request merging
merges two adjacent requests to one request
* Elevator
orders the requests to minimize seek times considering the physical data location (seek time is the duration of a head movement to reach an addressed sector). A "Prevent starvation" functionality is required, which ensures that requests which are not in the 'seek reducing' order are served in an adequate time frame. Here the various I/O schedulers have different strategies.
Available I/O schedulers
Noop scheduler
The noop scheduler implements only the request merging function.
Deadline scheduler
The deadline scheduler implements request merging, the elevator and in addition tries to prevent request starvation (which can be caused by the elevator algorithm) by introducing a deadline for each request. In addition, it prefers readers. Linux does not delay write processes until their data are really written to the disk, because it caches the data, while readers have to wait until their data are available.
Anticipatory scheduler ("as scheduler")
The anticipatory scheduler implements request merging, the elevator and in addition optimizes for physical disks by avoiding too many head movements. It tries to solve situations where many write requests are interrupted by a few read requests. After a read operation it waits for a certain time frame for another read and doesn't switch back immediately to write requests. This scheduler is not intended to use for storage servers!
Complete fair queuing scheduler ("cfq scheduler")
The complete fair queuing scheduler implements request merging, the elevator and in addition embarks on the strategy to allow all users of a particular drive about the same number of I/O requests over a given time.
Defaults
The default in the kernel 2.6 is the anticipatory scheduler. A Novell/SUSE SLES 9 installation overwrites this with the "cfq scheduler".
阅读(1138) | 评论(0) | 转发(0) |
