Archives of the TeradataForum
Message Posted: Fri, 20 Sep 2001 @ 18:58:21 GMT
<-- Anonymously Posted: Thursday, September 20, 2001 10:52 ->
Because of lengthening response times, I'm looking at new areas of our machine. From PMON virtual display, I have been noticing that we often have Disk busy levels reported at 60% average, 100% for high node and 0% for low node. Digging further and looking at the resource display, I have found that when the machine is busy, 60% of the system amp are at 100% and 40% are doing nothing. The specific amps that are doing the work do not change. If I open the AMP Disk Use by Processor graph, it is always the same nodes that are either 100% or 0% busy. I've had one of our NCR Pro Services guys tell me that this is a normal pattern. Is anyone else experiencing the same thing? Why don't the 40% of the Amps at 0% ever do any work? Is there anything I can do to get the work spread further?
1. AMP usage is very low - 10-20% busy and CPU workload is very evenly spread. At the moment average 5.4 %, high 7.4% and low 4.7%
2. Data in large tables has skews of 3 or less.
3. Actual workload is changing continuously and does not appear to affect the disk usage. This includes time while running fast and multi loads, archives, etc.
Any information or suggestions will be appreciated.
|Copyright 2016 - All Rights Reserved|
|Last Modified: 28 Jun 2020|