SAVEFD is not about speed of access to data on the CMS minidisk, it's about
saving real storage:
- without SAVEFD each user accessing a common minidisk has his own copy
of the file directory
- with SAVEFD all users accessing that common minidisk share a common copy.
Since file directory entries are 64 Bytes each accessing a minidisk with 1000
files requires about 64kB in a user's virtual storage. This means that if you
had 1000 users accessing that minidisk concurrently about 64MB of main storage
could be saved by using SAVEFD.
Back in the years when SAVEFD was introduced, when IBM internal PROFS systems
often had 1000-2000 concurrent users, and when storage was still expensive and
correspondingly scarce, the use of SAVEFD often made the difference between
good performance and absolutely catastrophic performance with the paging system
thrashing and nobody getting anything useful done.
But on VM systems where the main load is caused by Linux guests and you have
only a few dozen (or even hundred) CMS users you can forget about SAVEFD. All
the more so if you stopped long ago thinking in MegaBytes when talking about
storage ..
Eginhard Jaeger
- Original Message -
From: Gentry, Stephen
To: IBMVM@LISTSERV.UARK.EDU
Sent: Thursday, May 06, 2010 3:54 PM
Subject: [!! SPAM] SAVEFD
With the speed of z9's and z10's and the speed of DASD subsystems, is there
much of a need to use SAVEFD anymore?
I guess if a minidisk had extremely high utilization then it might.
Thoughts?
Thanks,
Steve