Archives of the TeradataForum
Message Posted: Wed, 09 Apr 2003 @ 11:45:30 GMT
Thanks for the quick response. However, I could use some further clarification.
Your explanation about using more CPU resources on the node running arcmain process makes perfect sense to me, and from a ResUsageSPMA data perspective, that is exactly what I am seeing.
However, from a DBMS perspective, I was surprised that the AMP Usage was so skewed. Especially since the AMP IO was almost perfectly parallel. This implies to me that at least one AMP was performing significantly more AMP related work than all of the others, even though it was not doing any more IO work.
I know that the ResUsageSPMA data gives me a view into all work being performed on a node, regardless of whether it is done by processes serving the Teradata DBMS software or other processes, like arcmain, or any other OS process. The info from DBC.ACCTG only accounts for work performed on behalf of the AMP vproc processes. So I am still wondering what AMP specific work is being done by the skewed AMP(s) that are causing it to consume so many more CPU cycles than the other AMPs, even though all of the AMPs are managing roughly the same amount of IO.
Thomas F. Stanek
|Copyright 2016 - All Rights Reserved|
|Last Modified: 27 Dec 2016|