Portal performance

[Former Member]

Hi folks,

i do have several customers that have performance problems with their enterprise portal.

But these problems don´t have to be "real" problems with the portal itself. It is also possible that they are based in connection to the backend system, they can be network problems, slow database(s), the j2ee engine etc...

Do you know a tool or a process that will enable people like me to find the real problem systematically? Do you have an idea what to test first, have a look at what kind of logs to get some information about the performace bottlenecks?

The best case would be a unified process that will check all possible bottlenecks systematically (does not have to be automated) and help the guy doing the measurement in finding the solution. Is there anybody who already has a solution for this problem?!

I hope you can help me

Kind regards

Michael

[Former Member · Accepted solution]

Please have a look at the SAP Solution Manager Diagnostics (SMD). This tool has the ability to look at some performance characteristics of your J2EE Engine and Portal.

[Former Member · Accepted solution]

Ist der Solution Manger eigentlich mit weiteren Lizenzkosten verbunden?

[Former Member]

Hi,

we faced also that kind of problems. How many server notes did you set up? And do you use https?

Best regards

Andrea

[Former Member]

Hi Michael,

did you get some more information about that topic. I'm trying to develop a "strategy" to test the portal and backend performance.

Did you find something like a check list or even a unified process? Please let me know if you got some more about this.

[Former Member]

Hi Micheal

I would like to add some more points:

1. First and foremost, identify the customer scenarios for which the performance is poor and get a clear idea about these sceanrios and make sure that these performance problems are reproducible or repetitive. As rightly pointed out by Joerg, working on all the scenarios will eat up time and money.

2. Get some statistical data such as the %CPU utilization, Memory utilization on different servers (WebAS, R/3 Backend, Databases etc.)from the customer's productive systems, if possible so that the issue could be identified as response time or memory etc. This data will be very helpful to foucs on the right component.

3. Try to have an environment (dedicated and controlled) where the customer scenarios could be simulatedwith differnt monitors enabled so as to identify the exact bottleneck.

Hope this is helpful.

Thanks and Regards

Madhu

[joerg_nalik]

Hi Michael,

I can feel with you and your customers as I have had very similar experiences. There might be no ultimate solution since every customer case has its individual challenges. However, here some tips:

- process/methodology: most important! If there is no clear concept for your approach likely a lot of time and money will be wasted. The strategy which worked best for me is start with small simple tests first, like do a manual 1 user test and use all available monitors for OS, databases, back-ends, network, back-ends to get clues. The work your way up to load tests. More is described in https://www.sdn.sap.com/irj/servlet/prt/portal/prtroot/com.sap.km.cm.docs/documents/a1-8-4/how to perform sap enterprise portal load testing.zip

One thing to add beyond this paper is that you want to test individual systems, like back-ends first for performance by themselves before moving on to EP embedded scenarios. You also want to start with LAN tests and then move to WAN tests. WAN tests can be challenging to do but are often important to see the "end-user" true experience.

- tools: all available monitors, Mercury Loadrunner for load testing.

- network: In particular if end-users are sitting on other continents network latency time becomes often the dominant response time contribution part. Do two things: Reduce number of application roundtrips as much as possible (see above paper for details) and get the network IT people into the boot to see what they can do from their end.

All the Best, Joerg

[Former Member]

Hi Michael,

first you can check some things in the monitoring service (reply times, memory usage and so on). If you want to make a loadtest I would suggest using a tool like Mercury Loadrunner. I think it would be the best if you prepare some test cases and then use such a tool to find out where the bottlenecks are. Mercury should have adapters to get performance data from the OS, the portal and the backends.

Best regards,

Jens

[Former Member]

Hi Jens.

which monitoring service you mentioned ?

Is it a tool inside Portal ?