Re: Running a repository out of RAM?

* Troy Curtis Jr (troycurtisjr@gmail.com) [070617 15:21]:
> Well server load and network isn't really the issue. It really seems
> like I/O is a big reason.

It seems like you may be skipping over a few well-known reliable steps
and right into the realm of the experimental. If Linux's builtin
caching isn't doing well enough for you, then maybe I really could shake
a stick at your RAM =)

Seriously, though, your first step (IMO) should be RAID. After that,
measure performance on different filesystems. If it's still no good,
you should consider a dedicated system, so that it can use all available
RAM just for the buffer cache. Only if that's not an option would I
consider trying to implement some sort of persistent mmap scheme.

I don't think I've really said anything new here; other posters have
already given these same suggestions. I just hoped to point out that
you should pursue the more well-trod avenues first, and that "mmap"
might be a good keyword to add to your searches.

Vineet

-- 
http://www.doorstop.net/
-- 
http://www.aclu.org/		It's all about Freedom.