On 11/6/2015 2:03 PM, michelbutz wrote: > Thanks I understand that in the beginning SRB were quick and dirty however > now they have become more robust > If you can maybe guide me to the control blocks associated with them the same > way TCB is associated with a task > > I am thinking of WEQ WUQ and SSRB
Short answers [thanks for the kind shout-out Dave] from a lurker... - I'm "pretty sure" the WUQ and its member blocks (WEBs) are OCO. There might be a strict subset documented in the pubs. - Preemptible SRBs (the class with 3 types that GA'd in 1995) all were of the "can do long-running work" variety... "preemptible" meaning that the dispatcher can pre-empt them in favor of work with higher dispatching priority. The stern "quick and dirty only" warnings around local and global SRBs were largely because they were non-preemptible... once the dispatcher gave it a CPU, it had the CPU until it relinquished control voluntarily, no preemption allowed. If you want more background on preemption Michel, look for CMG/SHARE papers from Bernie Pierce. DB2 V5 (the initial exploiter of client SRBs) used them specifically to run queries intensive enough that the optimizer chose to parallelize them (package + plan permitting, of course). - SRBs (preemptible or not) pre-date ZIIPs by years. I believe from ZIIPs day 1 they could run any preemptible work unit (SRB or task). John Arwe z/VM OpenStack enablement ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
