Re: Informix vs. Sybase vs. Oracle vs. (gasp) MS SQL Server

From: David Williams <djw_at_smooth1.demon.co.uk>
Date: 1997/11/27
Message-ID: <6lorlDACEff0EwkT_at_smooth1.demon.co.uk>#1/1


In article <yut67pfc6qb.fsf_at_mew.corp.sgi.com>, Pablo Sanchez <pablo_at_sgi.com> writes
>>>>>> "David" == David Williams <djw_at_smooth1.demon.co.uk> writes:
>David>
>David> page level locks give HIGHER PERFORMANCE yes BItUT the reduce
>David> concurrency and hence LOWER RELIABILITY.
>
>How do you assert "reduced concurrency" when the unit of
>measure is tpmC: transactions per minute of type C's?
>
> Informix... 24,000 transactions per minute of type C
> Sybase..... 34,000 transactions per minute of type C
>
  These can't have been from clients both updating rows on the same page   at the same time..
>I believe your assertion is wrong.
>
>David> most of the time people go for reliability over a little more
>David> performance. 90% of performance tuning is
>David>
>David> a) reducing disk I/O by only selecting/updating the minimum
>David> amount of data.
>
>agreed
>
>David> b) database schema / index usage.
>
>agreed
>
> c) proper hardware for both the client and the
> server
> d) configuring the server properly:
> - log on its own device (sequential writes in
> nature)
> - data on its own set of devices due to its
> mostly random access nature for OLTP
> e) proper network bandwidth and proper use of
> packages
> f) ...
 

-- 
David Williams
Received on Thu Nov 27 1997 - 00:00:00 CET

Original text of this message