Re: Theory of Timeseries extensions to SQL and database
From: Pablo Sanchez <pablo_at_dev.null>
Date: 27 Oct 2002 22:30:51 -0600
Message-ID: <Xns92B4DBD0E880Dpingottpingottbah_at_209.189.89.243>
>
> I've seen this proposed as a design several times over the years
> (Sybase IQ
> server used a flavor of 'vertical partitioning' too, and there are
> a couple of systems out there (required technology?) that make
> extensive use of clever compression schemes), but every time
> someone has gone this way the resulting system has either sucked on
> updates,
Date: 27 Oct 2002 22:30:51 -0600
Message-ID: <Xns92B4DBD0E880Dpingottpingottbah_at_209.189.89.243>
paul_geoffrey_brown_at_yahoo.com (Paul G. Brown) wrote in news:57da7b56.0210271944.343af261_at_posting.google.com:
>> In effect, all columns are indexed.
>
> I've seen this proposed as a design several times over the years
> (Sybase IQ
> server used a flavor of 'vertical partitioning' too, and there are
> a couple of systems out there (required technology?) that make
> extensive use of clever compression schemes), but every time
> someone has gone this way the resulting system has either sucked on
> updates,
But isn't that always the case? "You can't have something for nothing" type of rule? Or more succinctly: a trade off. I wonder how long it takes google to run updates on its indexes. I'm assuming they must have a constant index build going on a mirror and/or they're doing some incremental adds to their index.
-- Pablo Sanchez, High-Performance Database Engineering http://www.hpdbe.comReceived on Mon Oct 28 2002 - 05:30:51 CET