"Gurvinder" <gk@gcol.com> wrote in message
news:17e62da8.0401300929.6013c53d@posting.google.c om...
> I am using a solaris (2.7) workstation (multiple network interfaces)
> as a gateway for a private network. Users on the main network have
> occassional need to get access to the private network for downloading
> of data. The workstation is also tasked with other more higher
> priority tasks. I've set up the IP forwarding and routed correctly.
>
> My question is if I am downloading a lot of data (i.e. a few
> gigabytes) via
> the workstation, what load does this have on the cpu? Can the routed
> daemon be assigned a low priority. My data downloading is not a high
> priority, but needs to be complete when I come in the next day.
>
> I tried some stress tests, and in downloading ~1GByte of data with no
> major tasks running on the workstation, the perfmeter CPU reading
> shows very little
> activity - around 5%. I can't beleive the CPU isn't being loaded more
> with all those packets going through the interfaces.
>
> Thanks for respones.

Hello,

Great question!

If your really want in-depth profiling of applications on your
Solaris platform, both system-wide, and per-process, (user
selected processes for monitoring available only in the fully
licensed version), to find out not only extremely accurate
CPU usage, but also Memory usage, Network I/O usage,
Disk I/O usage, as well as a number of other system
paramenters & metrics, continuosly, over time, with an
extremely low-overhead, (measured significantly less than
1% over time), tool, then might I suggest checking out the
DPMonitor at www.deltek.us.

Typically, I/O datastreams are I/O intensive, not CPU intensive.

Load up the DPMonitor, (there is a 10 day free Trial
Evaluation License available, and the product is
downloadable off the Website), and map out exactly what
is happening on your platform, graphically. Graph data of
performance metrics are also exportable to CSV format
for use in other applications, if so desired.

DPMonitor product consists of a Performance Agent
that runs on the application Server being monitored,
(AIX, Solaris, Windows, and there is even an Oracle RDBMS
Agent), and the data is captured locally and sent via network,
to a GUI Performance Monitor Console process, called the
Performance Explorer, that runs on a Windows workstation.
This Performance Explorer consle manages Agent data from
all Agents, crunches numbers, displays auto-scaling, coloful,
dynamic graphs of all metrics, and you can event set up
Probes, to check for certain conditions or thresholds,
and take actions based on those thresholds or conditions.
Well worth the 10 day Trail Evaluation

I have used this product with tremendous success in numerous
problematic application platform situations. Check it out.

Regards,
Scott Richardson
Sr Systems Engineering Consultant
Marlborough, MA USA