The way I've always understood this is that SDSF schedules an SRB into the target address space to get spool data in buffers.
The difference between "S" (SDSF Browse) and the "SB"/"SE" (ISPF Browse/Edit) line commands is that with the latter commands, all the data is probably fetched up front. With "S", SDSF gets the data only when you need it, as you need it. If you never scroll down to the bottom of the data, SDSF never needs to get the data from the target address space. Further, the SRB that SDSF schedules to the target address space does not force the job to swap in. This means that if the low priority job happens to be swapped out, the SRB won't run. SDSF designers don't want to give a low priority job a "benefit" (forcing the system to swap it in) just to view the in-memory spool data. Finall, I have always understood that the SRB "times out" after 10 seconds. Thus, the delay. If the job is swapped in, the SRB will run "instantly", but if swapped out, the SRB sits and waits until it times out. Perhaps the main difference here is that when you issue a FIND command with SDSF Browse, SDSF simply skips the SRB, so you're never subject to the 10 second timeout. The above is basically my understanding just by thinking about how this probably works. Someone who actually knows might chime in with a better answer. Brian On Thu, 22 Jun 2006 12:49:48 +0930, Fred Schmidt <[EMAIL PROTECTED]> wrote: >Have any of you noticed poor performance in SDSF when you use the SB or SE >action characters to view a running job's output? > >When I do a FIND in a job with only 666 lines, it takes at least 6 seconds >to respond. Compare that with subsecond response for the same FIND command >when using the S action character to view the same output. If you are >looking at a job with a significant number of lines of output, you may as >well come back after lunch! The initial entry to view the job's output >with SB or SE is also very slow. > >This poor performance only occurs when viewing the output from a job that >is running. For jobs that have ended, performance is normal. > >We see this on both z/OS 1.4 and 1.7. It doesn't make any difference >whether we look at job output from the DA panel or the ST panel. > >Regards, >Fred Schmidt ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
