Monday, October 25, 2010

Microsoft Analysis Service Performance 3

Now a new tip for the SSAS 2005.
  1. make attribute relationships
  2. remove unused aggregations
Then in any Normal Scenario [according to Microsoft], this will solve it unless you have the [Large Dimension Problem] then it will not affect anything.

the Large Dimension Problem starts when you have at least one dimension with 170,000 records and more. In my case I have 220,000 records So I guess there are no solution.

I read a lot about migrating from SSAS 2005 to SSAS 2008, and also from 32bit to 64bit. but the reviews still as it. it should enhance the memory usage, but the same processing.

another thing that might be interesting, when I browse from Excel 2007 @ the server things are mmm not that bad, but doing the same from a different computer, then the disaster happens when I get to the large dimension.

what Microsoft expect to have small dimensions for making them work fine !!!!!
Enhanced by Zemanta

Sunday, October 10, 2010

Microsoft Analysis Service Performance 2

After we got confused, why our cube is too slow when accessing it from other combuters in the same LAN or even when browsing the cube within the same server.

Things to check:
  • Make the server 64 bit
  • Install SQL server 64 bit
Config to check:
  • Use Windows heap instead of SSAS heap

MemoryHeapType from 1 to 2

and HeapTypeForObjects from 1 to 0

  • Disable the option of blocking other queries if one of them is taking longer time

CoordinatorQueryBalancingFactor from -1 to 1

and CoordinatorQueryBoostPriorityLevel from 3 to 0

This should be done in the msmdsrv.ini file

All this content is coming from Mr. Ashotosh sharma from Microsoft

but till now it didn't solve the problem completely!!! so we are focusing now in the design tips.
Enhanced by Zemanta

Thursday, October 7, 2010

Microsoft Analysis Service Performance

While the past 3 months, I used to develop an OLAP project using SSAS 2005. After developing and deploying it, without data it looks good. when running the ETL's to populate the data in it! the hell is coming up.
The cube begin to be slower each time we tried to view a dimension with 200K records!

Then we tried partitioning the data, and how could Microsoft did that. they were putting for you a SQL query to partition the data!!!!! why didn't they do it as in SSAS 2000.

another thing, when we deploy the same solution to SSAS 2000 the performance is amazingly high!!!!

So, we partitioned the data, and tried to make the Cube MOLAP 100% and again it is only 4%.
and we could not guess why, cuz it is just a wizard!!!!!!!

We are now trying to tackle this problem, as soon as we solve it. I will post the solution in here.

Enhanced by Zemanta