Re: Large databases

From: Douglas Lucy <dlucy_at_allegro.amf.com>
Date: Tue, 21 Sep 93 06:52:17 -0500
Message-ID: <hztnsAHGBh107h_at_allegro.amf.com>


In <9309171424.AA11176_at_bravo> kfulton_at_mgrm.com (Keith Fulton) writes:

>I've got a friend thinking about downsizing an enormous
>application from a 3090 to a Unix box, and he's thinking
>about using Oracle. I want to get y'all's reaction to
>this.

Almost all modern RDBMS'es have some form of multi-volume capability; i.e. one database of one of more files can physically span more than one disk. Oracle certainly does, so you can _basically_ ignore worries about how big a db file Unix can handle. Depending on the type of transactions your application will perform, you can install different types of RAID arrays to store different-use parts of the database to increase performance (RAID X yields higher performance for long, sequential reads, while RAID Y yields higher performance for short, random reads/writes, etc).

However, your application may need to create large temporary files, and now Unix can be trouble. (almost) the only Unix with 64 bits offer the ability to store single files larger than 2 Gb.

Another nice-for-downsizing benefit of moving to Unix is the very attractive combination of price and performance the new multi-processor Unix machines offer. A low cost Sequent with 16 processors running a version of your favorite database re-engineered for multi-processor Unix will provide some very good performance gains, while still spending very little on the hardware.

Of course, you'll have that big conversion and re-write- our-application-for-Unix-and-Oracle bill to pay.

-- 
Douglas Lucy
Allegro Consulting Group, LLC
dlucy_at_allegro.amf.com
Received on Tue Sep 21 1993 - 13:52:17 CEST

Original text of this message