The buffspecs parameter has no effect. It has been ignored starting way
back when TurboIMAGE/XL was first created, IIRC. It is my understanding
that the first DBOPEN immediately uses the maximum value for buffspecs.
So whatever setting you use is ignored.
Denys
-----Original Message-----
From: HP-3000 Systems Discussion [mailto:[log in to unmask]]On
Behalf Of Emerson, Tom
Sent: Friday, January 23, 2004 4:57 PM
To: [log in to unmask]
Subject: [HP3000-L] A DB puzzle for the weekend
something came up today that prompted me to think about this, and after
a bit of research, I found that some of the parameters of our production
database are "default" values -- it is pretty much a "given" that
"defaults are bad", but I'd like some clarification on the following
[and suggestions on how to determine the "proper" settings]
In particular, the parameter I'm talking about is the BUFFSPECS as
reported in dbutil:
Buffer specifications:
12(1/2),13(3/4),14(5/6),15(7/8),16(9/10),17(11/12),18(13/14),
19(15/16),20(17/18),21(19/120)
as I understand this, the system will allocate a dozen buffers for the
first two users, add one more for the next two, add another for the next
two, and so on until there are 20+ users (at which point no more buffers
are allocated)
Instinctively "this seems bad" because if there are lots of "quick
accesses", the number of concurrent opens might fluctuate from half a
dozen to 20+ and back again in the span of a few seconds -- every time
you cross a "user count boundary", the system has to either allocate or
free up "one buffer" -- I'm envisioning severe memory
thrashing/fragmentation
OTOH, once things "really get going" for the day [and considering
potential accesses via an Oracle "gateway" and/or the standard ODBC
interface], the number of "concurrent" opens might hang out around the
50-100 mark, at which point "21 buffers" seems far too small
am I right about this being, shall we say, "less than optimal", or am I
just being paranoid? What do people recommend for a more "realistic"
setting of "buffspecs"? [again, considering both oracle and ODBC access
methonds in addition to normal HP-based (cobol) processes using DBOPEN]
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
|