There is something reassuring in the fact that these kinds of performance numbers are available and looked at.
/Tom Kern On 8/31/2012 02:29, Jim Mulder wrote: >> I still like the >> concept of having IPL and NIP somehow resident on the SSD for very >> fast loading. Perhaps in z/OS 2.3 <grin>. > > I would recommend against spending any of our development > resources on that. The actual loading of code contributes > only a small amount of time to system initialization up through > Master Scheduler Initialization. > > Here is an example of IPCS's IPLDATA STATUS > for a z/OS 1.13 IPL on an EC12 processor: > > *** IPL Statistics *** > > IEAIPL10 00:00:00.000 ISNIRIM - Read SCPINFO > IEAIPL20 00:00:00.000 Test Block storage to 2G > IEAIPL11 00:00:00.019 Fast FIND initialization > IEAIPL31 00:00:00.000 LOAD service initialization > IEAIPL30 00:00:00.000 Load IPLWTO. Allocate IPL Msg Q > IEAIPL46 00:00:00.197 Read SCHIBs into IPL workspace > IEAIPL49 00:00:00.000 Process Load and Default parameters > IEAIPL50 00:00:00.357 IPL parmlib - process LOADxx and NUCLSTxx > IEAIPL51 00:00:00.000 System architecture > IEAIPL43 00:00:00.004 Find and Open IODF data set > IEAIPL60 00:00:00.000 Read NCRs from IODF > IEAIPL70 00:00:00.054 UIM environment - load CBD and IOS services > IEAIPL71 00:00:00.048 Build DFT for each device > IEAIPL08 00:00:00.000 Read EDT information from IODF > IEAIPL40 00:00:00.028 Read MLTs from nucleus > IEAIPL42 00:00:00.002 Read NMLs from nucleus (IEANynnn modules) > IEAIPL41 00:00:00.361 Read PDS directory entries and CESD records > IEAIPL05 00:00:00.000 Build and sort NUCMAP > IEAIPL02 00:00:01.403 Load nucleus modules > IEAIPL04 00:00:00.003 Allocate PFT and SQA/ESQA > IEAIPL14 00:00:00.000 Build LSQA/ELSQA for Master > IEAIPL09 00:00:00.015 IAXMI - PFT, master RAB, etc. > IEAIPL07 00:00:00.002 Update AMODE for nucleus resident SVCs > IEAIPL03 00:00:00.009 Build UCBs, ULUT, etc. > IEAIPL18 00:00:00.006 Copy and relocate EDT to ESQA > IEAIPL99 00:00:00.109 Page frame table and cleanup > > Total IPL Time: 00:00:02.628 <snipped....> > > IEEMB860 00:00:32.481 Uncaptured time: 00:00:00.079 > > Total Time: 00:01:45.504 > > > Note that loading the nucleus was only 1.403 seconds, and > CLPA was only 8.430 seconds. Out of the total of 105 seconds, > the largest contributors are initializing I/O devices (5.958 > for non-DASD, 41.393 for DASD), and Dynamic Pathing (24.896). > Together these acount for 69 of the 105 seconds. Loading > from SSD would have no effect on this. > > Jim Mulder z/OS System Test IBM Corp. Poughkeepsie, NY ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
