Quantcast
Channel: SCN : Discussion List - BI Platform
Viewing all articles
Browse latest Browse all 5403

How to size the Adaptive Processing Server in a Clustered Environment

$
0
0

Hi All,

 

Greetings for the day!!!!!

 

We are running with BO 4.0 SP6 version on Win 2008 server.

We want to do APS split, so gone through couple notes "1694041", "Best Practices for SAPBO BI 4.0 Adaptive Processing Servers" and other "Companion guides" to do this. We came up with good understanding, but when summing the min XMX (Java Heap Size) parameters for the various APS Servers (MDAS, DSL, LCM ..) from above mentioned sap note, the total amount of required RAM is 26 GB for a minimum optimized APS Split. But in our case we have only 16 GB available and want to use 10 – 12 GB on this and rest for OS and other things.

 

We read, that we can still do APS split..... with available memory for the services which we use regularly.

Based on that we did the below Split:

 

Data Access DSL: with --> Xmx4g (Java heap Size)

DSL Bridge Service , WebI Monitoring Service, Security Token Service, TraceLog Service

 

MDAS OLAP Analysis : with --> Xmx4g

Multi Dimensional Analysis Service, TraceLog Service

 

Data Access Other : with --> Xmx1g

Adaptive Connectivity Service, Custom Data Access Service, Excel Data Access Service, Rebean Service

TraceLog Service

 

Monitoring : with --> Xmx1g

Monitoring Service, Document Recovery, TraceLog Service

 

Process of Split:

 

  • Cloned original APS and stopped --> Started the new one (APS1) with all services --> Cloned the new APS1 for Data Access DSL (APS2), MDAS OLAP (APS3), Data Access other (APS4) and Monitoring (APS5) APS's--> Removed all the services and added required as mentioned above for the 4 cloned APS's--> Given java heap size 4GB, 4GB, 1GB, 1GB respectively --> Removed all the added services to these 4 APS's from APS1 --> started all the cloned APS's (APS1 to APS5).

 

Here is the problem, services are started and working fine, but we can see "Blue Broken Dot" Error in Health status for these cloned APS's and also for the other services which are already running in CMS.

  • Followed the mentioned note to trace the above error "1725208", but didn't understand well.
  • Here we found one interesting thing, all the cloned APS's services are mapped to the original APS (Obviously, cause these are cloned from it).

        But, When we click on Blue broken dot of a particular server, it navigates to Monitoring Tab , then --> watch list --> Select the server (The one             which is showing Blue broken dot) -->  Edit --> Caution rule --> Servers --> Metrics --> here it is showing the metrics of this service is mapped to         original APS. Also showing an error like "Possible mismatch in the name". cause we stopped the original one.

 

My concerns:

 

  • Are, we did split in a right way..?
  • If yes, why Blue Broken Dot error in health status..?
  • Are we really need to bother about this Error .?
  • If yes, Is that OK, if we re-map metrics to cloned aps's (Help me with process) ..?
  • If we need to run the original APS too to avoid this error, then what is the definition of the split..?, as we have already other aps's running with same services, best practices saying do not disturb original one, clone it and stop it.

 

Please let me know your thoughts on the above....

 

Also help me with best way to split APS with 16 GB of RAM (We want to use only 10 - 12 GB in this out of 16, rest 4 GB for OS and for other activities) to get best performance ...?

 

Our Primary focus : IDT, BICS, WEBi, Dashboard, LCM

 

Thanks in Advance .. 

 

Regards,

Raju


Viewing all articles
Browse latest Browse all 5403

Trending Articles