cancel
Showing results for 
Search instead for 
Did you mean: 

Reporting Agent for olap cache

Former Member
0 Kudos

Hi All,

Is it possible to create a reporting agent job to fill the olap cache for a query that has a variable popup?

Is there anyway to give variable input parms for reporting agent to fill the olap cache?

Thanks

Mike

Accepted Solutions (1)

Accepted Solutions (1)

Former Member
0 Kudos

Need to give some thought about where the data will be cached for the queries that you would run thru RA.

If you have multiple Appl Servers in your landscape, you may have a couple of options - if you cache to main memory, you would probably want to then schedule the query to run on each Appl Server so it is cached in each Appl Server's memory - dependings on if you have differnent groups of users logon on to different Appl Servers, or if they are just used for load balancing and users could end up logging on to any Appl Server.

A better choice might be to cache to a Cross Appl Server BLOB or Cluster table, or even flat file. Then the results can be shared across Appl Servers and the query only run once thru RA.

Former Member
0 Kudos

Does anyone know a way to write precalculated data to the OLAP cache and to keep this data in the cache.

I managed to write the data to the cache but during the day rerunning the query does not use the cache.

Thanks!!!

Former Member
0 Kudos

Nili - There are many factors that come into play - system settings, configuration of the OLAP cache, InfoCube and query settings, the OLAP cache settings, e.g. read mode, persistence settings, etc. Then there are issues with your landscape and evironment, appl servers, user assignment to appl servers, frequency of loads, and other things that could invalidate the cache contents. Variable settings come into play.

The queries themselves and the way users run them vs under Reporting Agent can be the issue, as well as the size the result.

Not to mention, of course, that there are also a few OSS notes that deal with OLAP cache invalidation.

So it's not necessarily a simple question to answer. A lot more information is needed.

Here's a couple resources on OLAP cache -

https://www.sdn.sap.com/irj/servlet/prt/portal/prtroot/docs/library/uuid/9f4a452b-0301-0010-8ca6-ef2...

https://www.sdn.sap.com/irj/servlet/prt/portal/prtroot/docs/library/uuid/ef0e1d95-0501-0010-b0b3-aed...

https://www.sdn.sap.com/irj/servlet/prt/portal/prtroot/docs/library/uuid/7c361b95-0501-0010-c0ab-897...

BW Expert has also had articles on using Reporting Agent to warmup the OLAP cache.

Here's a couple of things that would be needed to begin to try to pin down your problem.

What are the cache settings for the InfoCube and queries?

- in memory, Appl server, Cross Appl server

- read mode

Are variables being used in the query? Are the variables set for "Can be changed during navigation"

Is reporting agent running the query after all loads and change runs are complete?

How big is the result? If using Main Memory caching, according to an SAP presenation I have, if the result set exceeds approx 1/5 of the OLAP cache size, the result will NOT be written to the cache.

Are users running the query exactly as Reporting agent runs it? No extra drill downs, same restrictions?

Message was edited by: Pizzaman

Corrected the % of the result set size - originally thought the size limit was 5% of the cache, but in reviewing, SAP says if greater than approx 1/5 (or 20%).

Answers (3)

Answers (3)

Former Member
0 Kudos

Hi Pizzaman,

Could you please tell us, what is the optimum size of data for caching to Flat file, Cluster table and BLOB table. Which one is better of the 3 options.

Thanks,

Srinivas

Message was edited by: SRK

Message was edited by: SRK

Former Member
0 Kudos

I have never really done specific testing to compare flat file vs the Cluster / Database options. I would like to see someone do that, but my personal preference is to use the database options for two reasons - space mgmt is going to be a little more resilient by storing in a table/tablespace that I think the DB offers more control over expansion. The flat file, once you fill the file system created for the cache, what do you do? Secondly, I think you have more DB buffering options that might work better than any OS based file buffering.

Can't find the document now, but the figure I have read on the Cluster vs BLOB was Cluster for < 1MB and BLOB for > 1MB. While I can't attest to any precise transition point, I think the notion of BLOB being better for large cacehd resutls is consistent with DB BLOB benefits in general.

Another factor that comes into play I think is whether you intend to use Reporting Agent to cache a large result (think BLOB) or just allow the user queries to be cached as they run them, creating lots of smaller results sets to cache.

Former Member
0 Kudos

Yes you can create variants for the input variables under the variants section in Reporting agent.

Thanks.

Former Member
0 Kudos

Hi Mike

In the variable editor you can flag the variable as "changable during navigation". With this flag set, you should be able to process the report with the reporting agent.

Brian