ASO Aggregations

Question

<p>Hi,  I'm having a difficult time find the right balancebetween aggregation, cube size and retrieval time.  I have anASO cube that has 3 years of data in it, totalling at 160MB. It has a total of 8 dimensions, 6 of which that containmultiple rollups.  So the data is obviously being aggregatedat many different points.  The problem is that if I go to setthe aggregation size higher this will help make the retreival timesfaster, but at the expense of making the cube much larger.  For example an aggregation of 400 MB will take 14.79seconds, where an aggreagtion of 1500 MB will take 7.66 seconds.   My goal is to get retreival times down to 1-3seconds, where the user would notice almost no delay at all.</p><p> </p><p>Does anyone have any suggestions on how to find a good balancebetween faster retrievals and the different aggregation sizes? FYI, I've also tried changing whether members are tagged asDynamic or Multiple Hierarchies enabled, but I didn't find anynoticeable difference.  Is there anything else that I shouldconsider to making the retreivals faster?  </p><p>Thanks in advance.</p>

Answer 1

Given that disk space is relatively inexpensive, this should often be something you don't worry about as much as something like query performance. Especially since even a highly aggregated ASO cube is only a fraction of the foot print for a BSO cube. You should maximize as much aggregation as you can to get the best retrieval time possible and not worry about the disk space too much. Unfortunately in current releases of Essbase you cannot customize aggregations. You can use query tracking to improve aggregations by focusing on the areas most often queried. So I would suggest enabling that, let your users get in there and then start to aggregate again based on the results of the tracking. Of course remembering that by its very nature ASO cubes are very dynamic and some queries are just going to take a little longer. Something to also consider is how complex the query is and is the amount of time it takes to retrieve appropriate. Are you pulling back thousands and thousands of members? If you are then you have to expect a certain amount of time just to bring over the meta-data. Try turning on navigate without data. If your queries still take a long time to come back even though you are not pulling data, then it just means you have a large result set coming back and that's just the way it is. Also look if your result set is returning members with member formulas. MDX formulas can take a little while to run depending on how complex they are and how well they are written.<BR>