Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!usc!samsung!shadooby!sharkey!itivax!dhw
From: dhw@itivax.iti.org (David H. West)
Newsgroups: comp.sources.wanted
Subject: Re: soft-overflow ramdisk for MSDOS?
Message-ID: <4541@itivax.iti.org>
Date: 30 Nov 89 19:49:02 GMT
References: <4482@itivax.iti.org> <720017@hpsad.HP.COM>
Reply-To: dhw@itivax.UUCP (David H. West)
Organization: The Forgotten Legions of ... um ... er ...
Lines: 22

In article <720017@hpsad.HP.COM> walter@hpsad.HP.COM (Walter Coole) writes:
|You could get many of the features you mention by using a cacheing program, of 
|[...]   The main limitation of caches is that since DOS can't run a synching 
|process, writes can't be cached.  Typically one reads far more than one writes,
|so this isn't a major disadvantage.
 
 [mentions 12X performance improvement in upspecified application]

For my intended application (a /tmp directory), files are typically
created, read once, then immediately deleted.  This means that a
cache can save no more than half the disk accesses unless it delays
writing long enough to see the delete.  I don't know whether lazy
writing is common, but I suspect not.

I'll probably end up using a cache anyway, since 
- one came free with my computer (but it uses immediate write-through)
- no-one has pointed me to a soft-overflow ramdisk
- I don't have time to hack on ramdisk code. 

Thanks to all who responded.

-David          dhw@itivax.iti.org