cancel
Showing results for 
Search instead for 
Did you mean: 

Long Roll Wait time during RFC calls

0 Kudos

Hi Experts,

This is my statistics : only 0.2% of the 100 000 RFC daily calls are longer than 1000 ms specifically during JCO_USER calls from the portal reports. This behavior is observed irrespective of reports which rules out application specific suspections.

When I analyse my worload statistics, I see long avg roll wait time which is 90% to 100 % when compared to the avg response times.

When I see the detailed rfc trace using the report RSRFCTRC, i see some errors as follows.

Error RFCIO_ERROR_SYSERROR in abrfcpic.c : 3425

Error RFCIO_ERROR_NOHANDLE in abrfcio.c : 3111

Error RFCIO_ERROR_SYSERROR in abrfcpic.c : 1464

Error RFCIO_ERROR_SYSERROR in abrfcpic.c : 3268

Error RFCIO_ERROR_MESSAGE in abrfcio.c : 1660

Error RFCIO_ERROR_SYSERROR in abrfcsys.c : 6286

Does this have any connection with my issue? as all the users associated with these error entries are not JCO_USER!!!

Is there any other direction to solve my issue?

Thanks and Regards

Karthick Gopi

Edited by: KARTHICK GOPI on Mar 25, 2010 10:55 AM

View Entire Topic
Former Member
0 Kudos

Hi Karthick,

I believe these RFC messages are not associated with longer roll wait time. As you might know that the roll memory area holds the user context data if roll buffer & extended memory are exceeded.If the action performed by the user (in you case calling BI reports) takes longer time to render results then the roll wait time will be longe thereby increasing the total response time (Avg response time X total steps). This is a typical performance issue. You need to analyse what are the different times that constitute the total response time. Dissect by first looking at the Database times i.e look at the SQL time (BI report will ultimately run SQL query only, use ST04). Subtract the DB time from the total response time. You get a time whch consitutes UI rendering time + Network time spent between UI and Application server + Processing time in the AS + Network time spent between AS and DB.Plus some other times. For deeper analysis you can use STAD.The statistical records collect information individually for each transaction step such as response times, database times, network times, wait times, front-end times and more.

This tool will help understand in detail where the performance spike is located by analyzing the transaction activity step by step. Information like how many database records were selected, updated or inserted and in which database tables, what program name was executed, what screen name and screen number was called and so forth.

As far as your question goes your analysis seems to be in a correct direction but do remember that BI reports put lot of load on the system and of there are millions of records the database times will be high thereby increasing the avg roll wait time.

Thanks,

jyotish