On Sat, 2006-07-08 at 21:24 -0500, Ron Johnson wrote: > Rich Johnson wrote: > > I also notice that there's a big discrepancy in the readahead. > > What are the tradeoffs of bumping this number? It just preloads > > the disk controller's cache, right? > > Good question. It is the difference between the 2.4 auto-optimizations and the 2.6 auto-optimizations. There are a number of reasons they are that way in the first place. And it is not really the "controller's cache" as most IDE controllers don't have much IF ANY onboard cache. It is the kernel's read-ahead for scatter and gather. It doesn't hurt to try to bump it up, unless you have a fixed controller or something similar it should change. I have had some controllers completely LOCK a machine, if that parameter is changed during a disk write operation. Particularly, early versions of piix and cmd640x. -- greg, greg@gregfolkert.net The technology that is Stronger, Better, Faster: Linux Use Debian GNU/Linux, its a bazaar thing NOTICE: Due to Presidential Executive Orders, the National Security Agency may have read this email without warning, warrant, or notice, and certainly without probable cause. They may do this without any judicial or legislative oversight. You have no recourse nor protection.
Attachment:
signature.asc
Description: This is a digitally signed message part