On Wed, Apr 30, 2008 at 3:42 PM, Rene Herman <rene.herman@xxxxxxxxxxxx> wrote: > > On 30-04-08 20:07, Bizhan Gholikhamseh (bgholikh) wrote: > > > > My understanding is the readahead algorithm causes kernel to read more > data ahead of the time from ide device, this will improve the > > ide access, however, in our case we find out it is also impacting over all > system performance. Is there similar thing for write access for > > Linux 2.6.11? > > > > Okay, if you're as dumb as me, this is an extremely amusing question. What > would you have the system write ahead? > > Rene Come on Rene, the opposite of read ahead is write behind. ;) And yes the kernel has that and so does some controllers and most disks. So you have: kernel disk caching and the various elevator implementations which reorder writes to optimize disk seek activity. 2.6.11 should have a few elevators to choose from. The disks have caches as well, but in general they do not reorder so you just have a enable / disable choice. With Sata II you have NCQ on some disks. That is a cache that also reorders the writes to optimize disk seeks. I don't think NCQ is supported way back in 2.6.11 You have to be very careful to have well implemented barriers in the FS code to ensure journals etc. offer the protection they are designed to offer even in the presence of NCQ re-ordering writes. 2.6.11 did not have those barriers fully in place I don't think. Greg -- Greg Freemyer Litigation Triage Solutions Specialist http://www.linkedin.com/in/gregfreemyer First 99 Days Litigation White Paper - http://www.norcrossgroup.com/forms/whitepapers/99%20Days%20whitepaper.pdf The Norcross Group The Intersection of Evidence & Technology http://www.norcrossgroup.com -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx Please read the FAQ at http://kernelnewbies.org/FAQ
