April 10, 2013 at 12:00 pm
Hi,
Is there a way to limit the CPU usage of a process? When we process one of our SSAS database, the process (msmdsrv) the CPU is spiking to 85-90%, causing the rest of the processes to be unresponsive, this include one of our SQL 2012 server (on the same box).
I had a look at WSRM, but unfortunately this is doing the opposite I need to do for CPU. The target is the percent of available CPU bandwidth that is guaranteed as a minimum to the process group.
Any help will be appreciated.
Rem
October 24, 2013 at 12:58 am
Any luck on finding a solution? I have this problem when I process a cube, when the cube has data that is partitioned by month. Works fine if I'm processing 1 month, but if I process a whole year, then I'm guessing there is a problem with SSAS threading when multiple CPUs need to write to multiple partitions.
October 24, 2013 at 7:34 am
What I did is moving the building of the cube on a different server. SO when the process is creating the aggregate, the high CPU usage doesn't affect the data server.
Another thing that I found was to build the dimension, data-measure and indexes in different steps, to minimize the parallelism.
I don't know if this is the best approach, but it's where I stand for now. I am open to other ideas/suggestions.
Remy
October 24, 2013 at 7:46 am
Thanks for the info. We are going to try to use a seperate server as well.
For troubleshooting, we stripped out the majority of our partitions and had the cube process only 1 dimension, and still ran into all CPUs being used up. Doesn't make any sense, its not that much data.
Anyways, I'll try your other solution as well. thx
Viewing 4 posts - 1 through 3 (of 3 total)
You must be logged in to reply to this topic. Login to reply