Parallel Oracle NCR3600
From: Carl Ricketts <carl_at_churchill.co.uk>
Date: Tue, 26 Jul 1994 10:46:47 GMT
Message-ID: <CtJnA0.n47_at_churchill.co.uk>
Date: Tue, 26 Jul 1994 10:46:47 GMT
Message-ID: <CtJnA0.n47_at_churchill.co.uk>
Well, unless the system has changed radically from when I benchmarked the NCR3600 in San Diego using Oracle Parallel Server ...
The intraprocess comms goes via disk between the units (yep its slow) and this is the same for all Parallel Server implementations when you go between 2 boxes or Processing units - it goes via the disk. (This is termed Pinging) as the data block gets "pinged" from one unit to another. The trick with OPS is to minimise this Pinging.
Here at Churchill we have had 9-10 months getting this right (now on 7.1.3).
Have fun ! - ask the guys at NCR about our benchmark..........<grin!>
regards
carl
Received on Tue Jul 26 1994 - 12:46:47 CEST