Messages in this thread | | | Date | Sat, 20 Mar 1999 23:34:52 +0100 (MET) | From | Gerard Roudier <> | Subject | Re: disk head scheduling |
| |
On Sat, 20 Mar 1999, Linus Torvalds wrote:
> On Sat, 20 Mar 1999, Gerard Roudier wrote: > > > > The current request queue in Linux is probably a excellent helper for IDE > > disks and SCSI disks when using a small number of tagged commands and a > > small number of disks at the same time, but for high-end SCSI we want to > > use numerous disks at the same time and to queue them a large number of > > tagged commands. > > This is mainly a problem with the SCSI subsystem, rather than anything > else. The SCSI subsystem is not very good at feeding the requests to > different disks, much less different controllers.
I didn't notice that. Everything that comes from the request queue seems to be queued normally to devices. It seems it is the kernel IO-request queue size that may limit the queuing but not the SCSI subsystem when queues are configured large for controllers and devices. Indeed there is room for improvements in the SCSI sub-system, but it passes IOs to low-level drivers normally, I think.
> The basic IO-request queue is fairly simple, and it's certainly not > perfect, but most of the problems are actually at a different layer. One > obvious one is the SCSI subsystem that forces everything to be on one > queue.
The kernel IO-request queue is one queue and a device TASK SET is also one queue, the SCSI sub-system passes requests between the kernel request queue and device queues. The only limits, the SCSI sub-system cares of, are the maximum number of outstanding IOs, for each controller and for each device, that may suspend the queuing. But it still is the expected behaviour.
> Linus
Sorry if I missed something in your statements. Probably I did.
Gérard.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |