🚀 go-pugleaf

In article <dpn2.65.721007601@po.CWRU.Edu>, dpn2@po.CWRU.Edu (Damien P. Neil) writes:
|> In article <Bx9ME2.H40@voder.nsc.com> pad@galaxy.nsc.com (Paul Denny x8349) writes:
|> >As a starter I note that caching disc controllers
|> >are supposed to increase the speed of disc intensive
|> >applications but a salesman I spoke to 8^) claimed
|> >that these are flakey designs as a single bad bit
|> >can propogate through your data and corrupt it. He
|> >suggested additional system RAM and partition it
|> >in DOS to have a large cache area. Can anyone verify
|> >or debunk this and provide an explanation?
|> 
|> I do not have the good fortune to own a caching disk controller, but I have 
|> heard only good things about them. Perhaps the RAM would cost more than the 
|> controller, giving the salesman more profits? :-)

Hum.  A single bad bit can propagate through a main-memory cache
too.  I think he's just plain wrong.

He *may* have been thinking about the fact that some caching disk
controllers do delayed writes - that is, they don't actually write
the data to the disk when your CPU does a disk write, they hold the
modified data in the cache until they *have* to write it (for
whatever reason).  This is the way the Unix operating system caches
the disk, and it's *much* faster than the traditional DOS-style 
write-through cache, where writes always go immediately to the disk.

The downside of this is that if you lose power before the data gets
written to the disk, it's *gone*.  Most of these caching controllers
allow you to switch between delayed-write and write-through modes.
In fact, as I recall, the new version of SMARTDRV.EXE that is
supplied with Windows 3.1 knows how to do both modes.
-- 
Erik Horstkotte, Togai InfraLogic, Inc.
The World's Source for Fuzzy Logic Solutions (The company, not me!)
erik@til.com, gordius!til!erik - (714) 975-8522
info@til.com for info, fuzzy-server@til.com for fuzzy mail-server