3 Replies Latest reply on Jan 11, 2011 1:11 PM by smohr

    PITimeSeries Performance Metrics


      Hi Guys,




      How many calls can PITimeSeries service support per second or per minute (considering we're passing 5 different PI points/PI requests)?




      Is there any performance metrics for PITimeSeries service? We’re making PITimeSeries service as an enterprise service in our landscape. Therefore, I want to understand the limitation (if any) of this service in advance.




      Our end to end call look like SAP System --- SAP PI Middleware ---PI Webserver. So we have to consider integration time also.









        • Re: PITimeSeries Performance Metrics
          Ahmad Fattahi



          Can you elaborate some more on the type of the calls you would make using PI Web Services? Talking specifically about limits would heavily depend on the nature of the calls and how resource-consuming they are.


          As a general comment on performance monitoring for any service you can use PI Performance Monitor Interface (or PI PerfMon) to track how a service is consuming memory, CPU, etc.

            • Re: PITimeSeries Performance Metrics



              I am calling PI web-service from SAP PI. Currently I am only using PIGetArchieveData to retreive data from PI server. The prototype is working fine also. However, I the performance metrics of the service is still not available. We are exposing this service as an enterprise service in our SOA landscape. So this service can be use by other business process for getting meter reads.So I am looking kind of information so that I can be aware what is the maxmimum call limit and what is the min time taken to execute ( considering hardware and message payload).


              Please let me know if you have inputs of this ( based on your testing and so on). As of now we haven't completely developed this service so we have to know its limitation too ( in terms of performance ).









            • Re: PITimeSeries Performance Metrics

              We are currently in the midst of performance profiling and improvements.  Much depends on your environment.  Integration time will likely dominate.  The situation you describe should support on the order of hundreds of calls per second for a fairly ordinary server dedicated to the web service. The number of events should be in the low tens of thousands.  Please remember, these numbers are very rough and depend on the usual factors -- memory, CPU, load, etc.


              The PI Web Service can scale using web farms.  You can also use the netTcp binding per the documentation to get a roughly 30% performance boost.  If you don't need Windows impersonation, you can drop back to the basicHttp binding and pick up some perf as well.