Messages in this thread |  | | | Date | Sat, 29 Jul 2000 12:54:00 -0700 (PDT) | | From | Andre Hedrick <> | | Subject | Re: ORBS Elevator... |
| |
On Sat, 29 Jul 2000, Rik van Riel wrote:
> On Sat, 29 Jul 2000, Andre Hedrick wrote: > > > It is one platter and a pair of heads. > > I will know the discrete track/sector arrangements. > > This is the first case where the "elevator" an be used as a strong boost. > > How about RAID1 devices?
I had a way to get cache table predictions but maybe another day. This a profile of the drive. But that ablity has been kill by the masses because I blew the presentation.
> In that case we have the data replicated over (usually) 2 > disks and we can optimise the reading of the data in such > a way that seek times on both disks are minimised, instead > of dividing the requests over both disks randomly and > hoping that the elevator code below can deal with it. > > Btw, if we want to do better IO clustering, we should probably > start doing IO clustering on *higher* layers. Currently we just > feed the elevator a random subset of the blocks that are to be > written and we hope that it's good enough... > > regards, > > Rik > -- > "What you're running that piece of shit Gnome?!?!" > -- Miguel de Icaza, UKUUG 2000 > > http://www.conectiva.com/ http://www.surriel.com/ >
Andre Hedrick The Linux ATA/IDE guy
- 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/
|  |