Newsgroups: comp.arch
Path: utzoo!henry
From: henry@zoo.toronto.edu (Henry Spencer)
Subject: Re: Sun bogosities, including MMU thrashing
Message-ID: <1991Feb1.164818.29577@zoo.toronto.edu>
Organization: U of Toronto Zoology
References: <5257@auspex.auspex.com> <3956@skye.ed.ac.uk> <PCG.91Jan18142616@teachk.cs.aber.ac.uk> <5390@auspex.auspex.com> <PCG.91Jan21160353@odin.cs.aber.ac.uk> <1991Jan21.225211.17757@gpu.utcs.utoronto.ca> <2006@necisa.ho.necisa.oz.au>
Date: Fri, 1 Feb 1991 16:48:18 GMT

In article <2006@necisa.ho.necisa.oz.au> boyd@necisa.ho.necisa.oz.au (Boyd Roberts) writes:
>The speed up was that you got 8 times more data per I/O.  Making the
>blocksize bigger speeds up I/O but at the cost of media wastage.
>
>Check out the BSD fast file-system paper...
>God knows why they added frags.  It
>would've been better to either bite the bullet or compromise on a smaller
>block size.  The complexities of that file-system far outweigh its advantages.

Check out the McVoy+Kleiman paper in the latest Usenix for the *correct*
answer to the performance problem:  don't mindlessly crank up the block
size, keep the smaller blocks but lay them out contiguously so the system
can read many at once.  Pity Berkeley didn't think of that.
-- 
If the Space Shuttle was the answer,   | Henry Spencer at U of Toronto Zoology
what was the question?                 |  henry@zoo.toronto.edu   utzoo!henry
