[THIN] Re: latency

  • From: "Braebaum, Neil" <neil.braebaum@xxxxxxxxxxxxxxxxxxxxxxxx>
  • To: "'thin@xxxxxxxxxxxxx'" <thin@xxxxxxxxxxxxx>
  • Date: Fri, 4 Oct 2002 13:43:54 +0100

> -----Original Message-----
> From: Bernd Harzog [mailto:bernd@xxxxxxxxxxx] 
> Sent: 04 October 2002 12:56
> To: thin@xxxxxxxxxxxxx
> Subject: [THIN] Re: latency
> Neil,
> Kevin Goodman (the founder of KevSoft, and the person who 
> invented and wrote TScale), believes that a substantial 
> portion of the kernel mode activity which occurs as the OS is 
> using the CPU to keep its own house in order is not reported 
> in CPU utilization counters.

I don't have any thing authoritative to dispute this, other than to say, I
can witness and monitor system activity - and this normally becomes a
requirement, when having issues along these lines.

> A specific example would be if a 
> kernel mode driver were to go into an infinite loop,

Yup - seen that sort of activity, had to trace it, and have managed to trace
it with perfmon and task manager.

> this 
> would block user mode activity and in turn cause CPU counters 
> to go down to zero,

Not in my experience.

The counters don't got to zero.

And besides - don't you think that's a little naive? Just because an
application runs largely in user mode, does not mean that all the hooks it
references do. Otherwise, things like task manager and / or perfmon would
never truly work.

> while the CPU would in fact be 
> unavailable to do useful work for users.

Threads can have priority, and / or be looping. In such case, it's the
kernel mode activity, that by it's very nature, can't (normally) be
interrupted (I'm think OS rings, here).
> We have observed the converse in many successful TScale 
> installations. Specifically, as TScale has eliminated page 
> file activity (which is kernel mode activity)

So is general memory usage - ultimately.

> from the 
> execution threads of applications, CPU utilization has gone 
> up. The reason it has gone up is that now more CPU is 
> available to do useful work for users, and as the CPU does 
> useful work for users the system performs better, scales better, etc.

The true reason is merely that there are possible less steps required to
serve the memory demands / and/or the requests can be served more quickly by
pages in physical memory (read: RAM) as opposed to having to fetch them from
virtual memory.

> The bottom line answer to your question is that yes, we 
> believe that CPU utilization counters are misleading because 
> they do not include CPU utilized by the OS in kernel mode 
> tasks.

I've *never* seen it, and I *have* had numerous instances where I've had to
trace looping (read: cpu draining) kernel mode drivers. As always, YMMV, but
I'm far from convinced that the counters mislead, or don't report - unless
the machine is too damn busy to respond to anything else, other than the
looping thread, in which case the box is effectively dead, anyways.

Not trying to pick a fight, just saw things mentioned that don't square with
me. Perhaps others may comment their thoughts?


This e-mail and its attachments are intended for the above named 
recipient(s) only and are confidential and may be privileged.
If they have come to you in error you must take no action based 
on them, nor must you copy or disclose them or any part of 
their contents to any person or organisation; please notify the 
sender immediately and delete this e-mail and its attachments from 
your computer system.

Please note that Internet communications are not necessarily secure 
and may be changed, intercepted or corrupted. We advise that 
you understand and observe this lack of security when e-mailing us 
and we will not accept any liability for any such changes, 
interceptions or corruptions. 

Although we have taken steps to ensure that this e-mail and its 
attachments are free from any virus, we advise that in keeping 
with good computing practice the recipient should ensure they 
are actually virus free.

Copyright in this e-mail and attachments created by us belongs 
to Littlewoods. 

Littlewoods takes steps to prohibit the transmission of offensive, 
obscene or discriminatory material.  If this message contains 
inappropriate material please forward the e-mail intact to 
postmaster@xxxxxxxxxxxxxxxxx and it will be investigated. 
Statements and opinions contained in this e-mail may not 
necessarily represent those of Littlewoods.

Please note that e-mail communication may be monitored.

Registered office: 
Littlewoods Retail Limited, 
Sir John Moores Building, 
100 Old Hall Street, 
L70 1AB 
Registered no: 421258 

This weeks sponsor 99Point9.com
99Point9 helps solve your unresolved technical
server-based questions, issues and incidents.

For Archives, to Unsubscribe, Subscribe or 
set Digest or Vacation mode use the below link.


Other related posts: