Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!agate!bionet!ames!amdcad!sun!pitstop!sundc!seismo!uunet!steinmetz!davidsen
From: davidsen@steinmetz.ge.com (William E. Davidsen Jr)
Newsgroups: comp.sys.ibm.pc
Subject: Re: Perstore and Maxtor
Keywords: controllers
Message-ID: <13154@steinmetz.ge.com>
Date: 16 Feb 89 15:27:51 GMT
References: <16761@srcsip.UUCP> <7981@netnews.upenn.edu>
Reply-To: davidsen@crdos1.UUCP (bill davidsen)
Organization: General Electric CRD, Schenectady, NY
Lines: 39

In article <7981@netnews.upenn.edu> silver@eniac.seas.upenn.edu.UUCP (Andy Silverman) writes:

| things like the clock speed of your computer.  If the interleave is too low
| (say 1:1 on an XT) the drive will try to send info faster than the computer
| can handle it and drive access slows incredibly as a result.  The trick is 
| trying to optimize interleave to get the highest throughput rate.

  Let me see if I can say that a little more clearly. When the CPU
requests reading of a sector, it is transfered off the disk, through the
controller and either into the CPU or memory, depending on whether a DMA
controller is being used (most AT's don't use DMA for hard disk). The
transfer rate for a sector is completely unrelated to the interleave
factor.

  After reading one sector, the CPU will do some processing and then
read another. This is where interleave comes in. If the interleave is,
say, 3:1, there are two sectors between consecutively numbered sectors.
While these pass under the heads the CPU can process the first read. By
the time the CPU is ready to read another sector the next sector is
almost at the heads, and the read takes place very quickly.

  If the interleave is is little large, one or more unwanted sectors
must pass by the heads before the desired sector comes along. This will
slow access by a bit. If the interleave is too small, the desired sector
will have passed the head already, and the disk must rotate one whole
revolution (or a little less) before the read can take place. This can
reduce transfer rate down to the floppy disk range.

  There is no "best" interleave, there are only compromises. If a small
interleave is selected (as low as 2:1 on an AT) programs will load very
quickly. Since there is very little CPU used in loading an image, the
CPU can "keep up with" the 2:1 interleave. That same interleave may slow
an application by a factor of four or five, because the data requires
more processing. Needless to say, the only solution which isn't a
compromise is a 1:1 track buffered controller.
-- 
	bill davidsen		(wedu@ge-crd.arpa)
  {uunet | philabs}!steinmetz!crdos1!davidsen
"Stupidity, like virtue, is its own reward" -me