IBM supplied a fix for this in the Latest DEC BRMS PTF SI57830.
26. In 7.1 and later, performance of save operations to large tape sets has been improved.
"tape exit 5 (the close that was taking more time due to number of volumes) and found a way not to do the updates if there was no volume switch"
On my Production LPAR, which is now at 33,000 sequences, the savings were around 15 minutes.
From: MIDRANGE-L [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of CRPence
Sent: Monday, August 31, 2015 10:57 AM
Subject: Re: BRMS nightly save time differences
On 31-Aug-2015 07:03 -0600, Steinmetz, Paul wrote:
High sequence numbers (57793) resulting when using append was the
420 per night, this media set dated back to March 2015.
I broke the media set, had Friday night's save start at sequence 1.
Sat and Sun appended as normal.
All three save times reduced by 20 minutes, equaling the once a week
Why would a high sequence number make a difference, it's only an index
in a BRMS database?
Maybe a question for BRMS support.
Physical positioning [of the tape heads] on the sequential [not random] media? Positioning to the physical row data, irrespective of sequential vs random\keyed I/O, would be unlikely to increase so precipitously; surely the time required for DB row-positioning could be measured in [milli]seconds.? Easy enough to test a CHKTAP specifying the larger sequence number, then another with a small sequence number, optionally with a rewind in between so as not to falsely inflate the timing of the latter.
This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list To post a message email: MIDRANGE-L@xxxxxxxxxxxx To subscribe, unsubscribe, or change list options,
or email: MIDRANGE-L-request@xxxxxxxxxxxx Before posting, please take a moment to review the archives at http://archive.midrange.com/midrange-l