Archives of the TeradataForum

Message Posted: Thu, 15 Nov 2007 @ 22:10:51 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

Besides caching the dictionary data FSG cache will hold frequently requested data. So you might get reads out of FSG cache the second time you execute a query which would improve response time.

There's also the Request Cache commented on by Bob Hahn. That would also improve response time.

With that said, I've never noticed either of these showing the performance gain that you mentioned. However, there is Full Cylinder Reads which could drastically improve the performance. The first run might not have gotten all/many Full Cylinder Reads slots it requested because of other workload on the machine (there's a limited number of slots for Full Cylinder Reads) where the second one might have been able to get more Full Cylinder Read slots.

However, if we are going to take workload into account then it could be any number of things that had an effect on the performance of your query. Instead of wall time you should be looking at the CPU seconds and the numbers of I/O's that the request used. That will give you a better idea of true performance of the query.

Robert Meunier

	<Prev	Next>		<<First	<Prev	Next>	Last>>