Re: AVAIL (EVENTS) Default Settings

From: John Brandon (brandon_at_dalsemi.com)
Date: 06/10/03 

Date: Tue, 10 Jun 2003 12:20:01 -0500

> Barry.Kierstein@HP.Com
> The values were based off of low-end Alphas, if memory serves me...

I take that to mean AXP 800 series. No wonder I red-line all the time with my
GS160!!! ;0

> The whole notion of "what are the correct values for the thresholds"
> is a complex one. It depends on the expectations on the machine, do
> you want to have some "reserve capacity" for unforseen workload
> spikes, or run the machine flat out? Even disks are different, as the
> variation of disk drives and what they are connected to make a
> difference. So, in short, after taking to the performance people
> here, making a template is a rather mindbending task.

I agree - "correct" values is very complex.

I understand that configuration, applications, type of hardware, firmware, O/S,
etc. all have their implications and play havoc with a standard. Back in the
VAX days we had the VUP standard - which has been replaced by the SpecInt and
the likes. Since most of these standards are out there, would it not be
possible to apply a SpecInt value to come to a default standard? Yes I
understand that it is not as easy or as simple as that...

However HP, Sun, IBM, and every other MFG is in the habit of comparing their
hardware against each other. Mine is bigger, better, faster, ... I believe it
is high time they start to deliver simple baselines of expectations for their
machines. Take the VUP, now make it a DPS for Default Performance Standard.
Go ahead and add all the disclaimers you want!

> However, there may be other ways to get at "optimum values". One
> thing, as a first step, is to include what the observed values are in
> with the event as well as the limit. This may or may not be best in
> the text of the event, but maybe other columns in the event pane.

Limit. That is the biggest "unknown" about performance. Even discussing these
issues with my VMS rep's bring that to point. It is always easy to compare the
performance to what you had (say an old VAX) to what you have now (a kick ass
Alpha). It easy to measure that. Then you also know you are pushing the limit
when users start to complain about delays or jobs that took 10 minutes now are
taking 20. You can only blame the network for so long before they start to get
wise! ;) Then it becomes "reaction mode".

> Anyway, all the values gathered from the node as well as the
> configuration data would be available. With this could be a dialog
> box from a right-mouse-click (double click brings up the node of the
> event) that would present the current threshold values, the current
> observed values, and an edit box for a new threshold value. Another
> would be to have a "this is really OK" option, and have the threshold
> value adjusted accordingly.

An automatic adjustment - based on observed work load - that would be a step in
the right direction. That automatic adjustment could be an application outside
of the main AM with the objective to build the default templates.

However it does not reflect the capability or limites of the machine. And you
 are right - mindbending! - however I present to you that there are products
out there (collection agents, performance advisers, and capacity planners)
that seem to have an established default (of sorts) for a whole mix of
configurations. Not that I am saying (or wanting) AM to perform these types of
duties - just stating that there seems to be rules-of-thumb already out there.

Maybe the automatic adjustment is coupled with a task to task application that
is checking its performance of file copy, IO transfer, network connection, etc.
during the automatic adjustment. Similiar to what the TCPIP NTP stack does for
time keeping standards.

> The above are suggestions to more quickly change values, but as far
> as the question of "what value is correct for me" kind of thing, that
> is tougher. I remember as a customer wondering about that, and
> learning as much as I could about various tuning philosophies, etc.
> The AM team is open for suggestions on this, as setting these values
> is key to making the tool work better.

So I am a dreamer - wait a tick - was not the GS series touted as the
DreamMachine??? ;)

Call me what you will, just do not call me Shirley!

John Brandon
VMS Systems Administrator
john.brandon.spam.me.not@dalsemi.com

Relevant Pages

  • Re: OMB to require standard Windows desktop configuration
    ... "The Office of Management and Budget will require agencies to move to ... a standard configuration for Microsoft Windows desktop by Feb.1, ...
    (comp.sys.mac.advocacy)
  • Re: Right. Own up.
    ... appliances you'd have a NEMA 14 configuration (two poles plus neutral ... since 120 is the nominal standard. ... We typically have, as I said, 127/220 in very old houses (AFAIK those are ...
    (alt.sysadmin.recovery)
  • Re: OMB to require standard Windows desktop configuration
    ... a standard configuration for Microsoft Windows desktop by Feb.1, ... Director for Management Clay Johnson, ... about securing Windows operating systems, ...
    (comp.sys.mac.advocacy)
  • Re: OMB to require standard Windows desktop configuration
    ... a standard configuration for Microsoft Windows desktop by Feb.1, ... Director for Management Clay Johnson, ... about securing Windows operating systems, ...
    (comp.sys.mac.advocacy)
  • Re: Destructor: not gauranteed to be called?
    ... >> documentation wether your expectations were justified. ... MSDN is a reference, just as a dictionnary or an encyclopedia. ... > would 'violate the C++ standard'. ... ToString is not part of the C++ standard! ...
    (microsoft.public.dotnet.languages.vc)