Path: utzoo!attcan!ncrcan!ziebmef!mcp
From: mcp@ziebmef.mef.org (Marc Plumb)
Newsgroups: comp.sys.amiga
Subject: Re: Disk Speeds Indeed
Message-ID: <1989Aug28.193641.19719@ziebmef.mef.org>
Date: 28 Aug 89 23:36:40 GMT
References: <7701@cbmvax.UUCP> <1555@convex.UUCP>
Reply-To: mcp@ziebmef.mef.org (Colin Plumb)
Organization: Ziebmef Public Access Unix, Toronto, Ontario
Lines: 24

In article <1555@convex.UUCP> swarren@eugene.UUCP (Steve Warren) writes:
>In article <7701@cbmvax.UUCP> daveh@cbmvax.UUCP (Dave Haynie) writes:
>>A2630, 4 Megs:
>>
>>	1024K	Reads 853333	Writes 503606 b/sec

>Awesome.  $?

I beg to differ.  Steve Beats gave me a figure of 1.2 MB/sec at the Exec
device driver level for reads from a CDC Wren III using a 2090A.  On
a 7.14 MHz 68000.  An awesome file system should be able to do at
least 1MB/sec on the same hardware on a 90% full disk.  1 MB/sec at all
is decent, but the true awesomeness of a file system comes when it has
to cope with full disks.

I don't know if he has the time, but I hear Steve is dreaming about something
better.  Does anyone else want to give it a go?  File system implementation
is, from my attempts at research, very poorly researched, but there is some
stuff on the Berkeley FFS, and "Disk File Allocation Based on the Buddy
System" offers some nice ideas.  Personally, I'm convinced of the
desirability of extent-based systems with background incremental disk
"optimisation".  But if you have a contrary opinion, go to it!
-- 
	-Colin Plumb