Archives of the TeradataForum
Message Posted: Wed, 24 Jul 2002 @ 13:55:25 GMT
| Subj: | | Re: UNION ALL |
| |
| From: | | Doug Drake |
Is there other concurrent workload on the system while you were testing? If so, then execution time is not expected to be a stable
metric.
How many times did you perform the first tests when you received the 12 seconds? It may have been possible that one or both of these
tables were in cache. Is it possible that any of the tables you were testing may have been bypassed due to synchronized scanning or
cache?
If the data types are indeed the same I suggest testing multiple times at different periods of the day possibly and measuring CPU and I/O
consumption. I would attempt to factor out the effects of caching, synchronized scanning, and concurrent workload.
Doug Drake
| 
