Hmmm. I hear you. (But the number of iterations is actually only 1/16, right? X'20000' is 1/16 of X'200000', right?)
Here is the data extrapolated out to a constant number of iterations. (The ratio for ESTAEX no SNAPTRC appears to be wrong in Jim's numbers.) Recovery Iterations CPU seconds Ratio ---------------- ---------- ----------- ----- ESPIE x'200000' 3.53 1.0 FRR x'200000' 45.66 12.9 ESTAEX (no SNAPTRC) x'200000' 1583.20 448.5 (28.0 is apparently wrong) ESTAEX (SNAPTRC) x'200000' 52648.96 14,914.7 Charles -----Original Message----- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Tom Marchant Sent: Friday, April 3, 2020 9:36 AM To: IBM-MAIN@LISTSERV.UA.EDU Subject: Re: ESPIE question (does ESPIE "cover" ATTACH'd sub-tasks) I saw the Ratio column. comparing lines 2 and 3 of the chart, the ratio doesn't make sense if the number of iterations for ESTAEX is only 1/10 of the number for FRR, yet it took over twice as long. Perhaps Jim will clarify. On Fri, 3 Apr 2020 09:12:56 -0700, Charles Mills wrote: >Look at the Ratio column for "normalized" numbers. > >ESPIE beats everything. That's the point. If (a.) all you need to trap is >program checks; and (b.) you expect a bunch of them -- use ESPIE. > >Charles > > >-----Original Message----- >From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On >Behalf Of Tom Marchant >Sent: Friday, April 3, 2020 8:10 AM >To: IBM-MAIN@LISTSERV.UA.EDU >Subject: Re: ESPIE question (does ESPIE "cover" ATTACH'd sub-tasks) > >The data presented shows that FRR is a lot better than ESTAE(X). >Perhaps you overlooked the number of iterations. > >-- >Tom Marchant > >On Fri, 3 Apr 2020 10:23:33 +0300, Binyamin Dissen wrote: > >>Interesting numbers. >> >>But I looked at the current doc and it still appears to be problem state only. >> >>Also, do you numbers include setup or just program check handling? I figured >>FRRs would be a lot better than ESTAE(X). >> >>On Thu, 2 Apr 2020 19:28:13 -0500 Jim Mulder <d10j...@us.ibm.com> wrote: >> >>:> These are my results from a benchmark I did 4 years ago: >>:> >>:>Testcases which loop recovering/retrying from an >>:>operation exception. >>:>Using default system trace size - 1MB per CPU, with >>:>20 CPUs, so 20MB of data to snap) >>:>z13 machine >>:> >>:>Recovery Iterations CPU seconds Ratio >>:>---------------- ---------- ----------- ----- >>:>ESPIE x'200000' 3.53 1.0 >>:>FRR x'200000' 45.66 12.9 >>:>ESTAEX (no SNAPTRC) x' 20000' 98.95 28.0 >>:>ESTAEX (SNAPTRC) x' 1000' 102.83 14,914.7 ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
