Path: utzoo!attcan!uunet!mcvax!kth!enea!maxim!prc
From: prc@maxim.ERBE.SE (Robert Claeson)
Newsgroups: comp.databases
Subject: Re: Corrupted database in Oracle
Message-ID: <624@maxim.ERBE.SE>
Date: 15 Mar 89 16:43:01 GMT
References: <2718@rtech.rtech.com>
Organization: ERBE DATA AB
Lines: 28

In article <2718@rtech.rtech.com>, billc@rtech.UUCP (Bill Coffin) writes:
> >From article <531@maxim.ERBE.SE>, by prc@maxim.ERBE.SE (Robert Claeson):
> > .... Under System V, Release 3.1 at least, there's a SYNC flag
> > for open() and fcntl() that sets synchronous write. 

> But that's only sysV.  BSD, I believe, still doesn't have this.

Most large databases runs under AT&Tix, I think, even though the Berklix
fast file system is better suited for this kind of applications (if it
wasn't for the lack of synchronous write).

> You can be clever about physically clustering pages.  This can produce 
> substantial performance improvements, since you can suck in an index 
> or sequentially scan a table much faster than on vanilla UNIX, where 
> pages can be scattered around any old place on the disk, causing longer
> seek times.  (Britton-Lee got a lot of mileage out of this ability).

I believe that the Berklix file system tries to keep the disk blocks
somewhat in sequence and the free list sorted. As always, I can be wrong.

Let's turn the page and see what AT&Tix 4.0 with the Berklix file system
and the promised transaction processing enhancements can do for databases
using the file system.
-- 
Robert Claeson, ERBE DATA AB, P.O. Box 77, S-175 22 Jarfalla, Sweden
Tel: +46 (0)758-202 50  Fax: +46 (0)758-197 20
EUnet:   rclaeson@ERBE.SE               uucp:   {uunet,enea}!erbe.se!rclaeson
ARPAnet: rclaeson%ERBE.SE@uunet.UU.NET  BITNET: rclaeson@ERBE.SE