write performance - broad generalization

From: Michael McMullen <ganstadba_at_hotmail.com>
Date: Fri, 3 Apr 2009 08:18:10 -0400
Message-ID: <BAY141-DAV7E50FFC180AF7EDD3252CA6890_at_phx.gbl>
Message-ID: <8BAFB96E2356436F9E425B0663105A73_at_vpmcm01>

Whenever I get a new server or even compare write performance between servers, I just create a 10GB tablespace (not temp). I time it and I think it gives me a kind of max write performance (for oracle), no other work is going on, just the one tablespace being created. Real quick & dirty test. Yesterday I got a brand new dev box and it took ~7 minutes for a 5GB datafile. A 10GB datafile on another box that I recently set up < 1 minute. Top of my head I think my new dev box is a dog. I don't need to go into specifics on the file system they gave me or the underlying storage. Just wondering if I'm way off base. I'll have to go through my docs. I think I've done test with multiple sqlloader sessions doing direct loading to see if my throughput matches my simple test but I'm not sure.  


Received on Fri Apr 03 2009 - 07:18:10 CDT

Original text of this message