Subject: Re: yamt-readahead branch
To: Chuck Silvers <email@example.com>
From: SODA Noriyuki <firstname.lastname@example.org>
Date: 11/17/2005 16:16:48
>>>>> On Wed, 16 Nov 2005 23:10:41 -0800, Chuck Silvers <email@example.com> said:
> I think the next most common case is that multiple threads read the
> file using the same file descriptor. for this, per-struct-file
> state doesn't help. the case where multiple threads read the file
> simultaneously using separate file descriptors seems relatively
I think the latter is more frequent in http servers and usual unix
applications than the former.
What the common case for the former case?
I guess database access is nearly random, and readahead doesn't help
(or maybe make performance worse even) for the database case.