Xref: utzoo comp.sys.ibm.pc:38733 comp.unix.xenix:8702
Path: utzoo!attcan!uunet!van-bc!jtc
From: jtc@van-bc.UUCP (J.T. Conklin)
Newsgroups: comp.sys.ibm.pc,comp.unix.xenix
Subject: Re: MAXTOR XT4170 with WD1007 Controller (& SpeedStore)
Keywords: MAXTOR XT4170 WD1007
Message-ID: <70@van-bc.UUCP>
Date: 25 Nov 89 00:14:51 GMT
References: <1187@otc.otca.oz> <10373@rls.UUCP> <1989Nov24.025335.20233@ddsw1.MCS.COM> <10375@rls.UUCP>
Reply-To: jtc@van-bc.UUCP (J.T. Conklin)
Organization: Wimsey Associates
Lines: 24

In article <10375@rls.UUCP> randy@rls.UUCP (Randall L. Smith) writes:
>In article <1989Nov24.025335.20233@ddsw1.MCS.COM>, karl@ddsw1.MCS.COM (Karl Denninger) writes:
>> NO!  Doing that makes the system quite a bit slower than if you leave it at
>> the real physical parameters!   You see, the controller has to remapping if
>> you set things up like that.  

>But Karl, that is the function of the WD1007-WA2.  Dos and earlier versions
>of Unix cannot (with native drivers) access those extended address ranges. 
>Since all the computing and remapping is done on the controller, this
>should not be slower.

Even if translation could be done instaneously, disk performance will
be seriously impaired.

The disk arm scheduling algorithm in the device driver has to know
about the physical disk geometry for optimum performance.  Typically
the scheduling algorithm assumes that track n is next to tracks n-1
and n+1.  A translating controller invalidates this assumption.

    --jtc

-- 
J.T. Conklin
	...!{uunet,ubc-cs}!van-bc!jtc, jtc@wimsey.bc.ca