Archives of the TeradataForum
Message Posted: Tue, 14 Dec 1999 @ 21:38:53 GMT
There might be a simpler answer: Teradata is designed to use all the resources available whenever it can, and when it is not otherwise told not to. So this CPU usage could be "normal."
If you submit a complex query, Teradata will automatically try to use 100% of the CPU resources available to return your answer as soon as possible. If you submit two equal queries, and you assume that they both CAN use 100% of the CPU, then the response time of the queries will no more than double running concurrently versus running stand-alone. However, in the concurrent case, you will usually see a less than 2x increase in response time due to efficiencies in the database engine. Please note this is for complex, all-AMP queries -- not simple, single-AMP queries.
For a complex stand-alone query not to use as much CPU as it can, you'll need to use something like the Fair Share Scheduler to segment off a portion of the resources as not available to that query.
Finally, I have found the Resusage data (for example, running the macro "Resnode") to be a much better indicator of how busy the system really is -- for Teradata dedicated nodes. You may be capturing some background housekeeping UNIX processes that go idle under a heavy query workload.
|Copyright 2016 - All Rights Reserved|
|Last Modified: 28 Jun 2020|