As usual response was incredible: fast, accurate, and useful. Not one
useless response out of about two dozen. Astounding. Everytime I
started to summarize, another response would come in.
In my original query I asked about increasing the performance of an IPX
through increased RAM, modified cache, and/or tweakable kernel
parameters, and where to read about such things. As was gently brought
to my attention, I neglected to specify my OS, so responses may apply to
1.x or 2.x.
On learning about performance the overwhelming consensus was Adrian
Cockcroft's publication(s) from Sunsoft Press. Two titles were
mentioned but I don't know if they are the same book: "Sun Performance
and Tuning", and "Solaris Performance Tuning". Also mentioned was
"Nutshell Handbook: System Performance Tuning".
On actually tweaking for increased performance the general opinion was
that it was debatable whether any significant improvement would be
worth the effort or the impact on current production. Though no one
felt there was much potential gain in tweaking them, the following
parameters were mentioned:
lotsfree, desfree, slowscan, fastscan, minfree, handspread,
maxpgio
I got numerous suggestions for determining the nature of current system
load and quite a few other suggestions for options to try. The
suggestions were too numerous to include here, but if anyone would like
to sort through them, I will be happy to forward the lot to you. I will
be hanging on to them for a long time. Suggested tools were:
top, vmstat, iostat, nice
Several hardware options were also suggested:
- if servicing NFS client calls, buy a PrestoServe S-BUS board
non-volatile NFS accelerator
- upgrade IPX sparc CPU chip with a higher performance Weitek
product @ about $1300.00 CANADIAN
- if CPU is loaded install a Weitek clock doubler chip.
1.5X performance and more just by replacing the CPU
- Weitek SPARC POWER uP chip
Max RAM on IPX: 64MB
Max SCSI on IPX: 3 S-Bus cards
My apologies to those wanted to give me more detail were it not for the
sparse detail in my query. You actually gave me exactly what I needed.
The IS (Information Systems (programmers) for those who asked) group is
at a point where they must decide upon a platform for their application.
Money seems to be their overriding concern and they apparently don't
want to spend any (they balked at buying a 1.05GB internal drive to make
a spare unused diskless workstation diskful). The responses all suggest
that they are going to have to come loose with some cash, and none
seemed even a bit exorbitant. I can't understand IS managers who won't
spend money on hardware until they're forced to, even though they
willingly spend much more in labor costs to figure out how to avoid the
hardware expense. So much for my soapbox.
Thanks to:
rao%hobbes.crc.com
gregr@cibc.com Greg Roberts
bismark%alta.jpl.nasa.gov
dwellbor@drao.nrc.ca Dale Wellborn
Poehlem%img.wdc.com Karl Poehleman
Birger.Wathne%vest.sdata.no
okenka%mail.msen.com
steve@pnw.opensys.com Steven Lee
poffen@San-Jose.ate.slb.com Russ Poffenberger
Kevin.Sheehan%uniq.com.au
pcaskey@swcp.com Paul Caskey
glenn@uniq.com.au Glenn Satchell
mark%lochard.com.au
coe%altair.becta.tscs.com Andrew Coe
anchi%starbase.neosoft.com
jdd@cdf.toronto.edu John DiMarco
strombrg@uci.edu Dan Stromberg
wwtz@chbs.ciba.com Wolfgang Wetz
I would have thought by now I had received all the replies but I
received another just minutes ago, so thanks also for any replies not
yet received.
Ross
ross.stocks@nt.com
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:23 CDT