Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site ttidcb.UUCP
Path: utzoo!linus!philabs!ttidca!ttidcb!speaker
From: speaker@ttidcb.UUCP (Kenneth Speaker)
Newsgroups: net.followup
Subject: Re: Unix from a snob's point of view!
Message-ID: <563@ttidcb.UUCP>
Date: Wed, 4-Dec-85 14:50:44 EST
Article-I.D.: ttidcb.563
Posted: Wed Dec  4 14:50:44 1985
Date-Received: Sat, 7-Dec-85 15:28:24 EST
References: <298@weitek.UUCP> <228@polaris.UUCP> <942@wcom.UUCP> <314@polaris.UUCP>
Reply-To: speaker@ttidcb.UUCP (Kenneth Speaker)
Organization: Transaction Technology, Inc. (CitiCorp), Santa Monica
Lines: 23
Summary: 

In article <314@polaris.UUCP> herbie@polaris.UUCP (Herb Chong) writes:
>In article <942@wcom.UUCP> frodo@wcom.UUCP (James Scardelis) writes:
>larger LOGICAL blocks, but you still end up working with them as collections
>of units of 512 bytes.  having larger logical blocks alone can improve system
>performance since you can make a bunch of I/O requests in a row.  i don't know
>if disk controllers typically used with VAX hardware understand chained
>I/O commands as i know IBM 370 channels do.
> 
>
>Herb Chong...
>
Actually the Unibus and Massbus both have DMA mapping hardware.  The system
then sets up the map to "gather" (on output) or "scatter" (on input) the
logically contiguous to physically dis-contiguous memory.  On the VAXen I
use with VMS, the cluster size used to write the paging file is 128 (512
byte) blocks, or 64K transfers.  On output it can always do this.  On page
in, it depends upon how many of the blocks which are "near" one another on
the page file can be utilized.  If about half of the blocks in a cluster
are needed, the other half can be mapped into the black hole, i.e., a single
page reserved for that purpose.

--Kne