Re: Microsoft destroys TPC-C records!

Shared nothing referes to system resources only. Versus shared everything or shared disk.. and so on. The trick is to minimize the network-traffic and the one there is, better be fast.
There are certain types of queries that are more applicable for parallelization
than others and it is also important to decide in how far it makes sense to e.g.
replicate data on versus nodes.
Scalability is not automatic. It depends heavily on the quality of the optimizer,
the DB schema, infrastructure and the nature of the query. E.g. an insert of one row will not run faster by adding more machines. A select count(*) from t on a nicely partitioned table however will scale very well.

Just my two cents
Cheers
Serge Received on Wed Mar 01 2000 - 00:00:00 CST