How to calculate bandwidth requirement

[Cameron Simpson]

On 13Nov2009 08:26, Stainforth, Matthew (SD/DS) <Matthew.Stainforth at gnb.ca> wrote:
| I think this is going to depend greatly on the maturity of your
| organization and could look like any of these:
[...]
| 4.  proper load testing tools in a dedicated PST environment 

If feasible, I would go for (4).

The tricky bit is making a test suite with your load test tool
that is representative of how things will be done in the real world.
Once you have that you can set up something to measure usage.

Remember that lack of bandwidth (up to a point) won't result in failure
for your clients, more probably degraded performance. The flip side of
that is that if you're testing on a LAN the high local bandwidth can
skew the results of your test (i.e. indicate you may need more bandwidth
than you really do) because I/O bound stuff can run faster. Make sure
your test has tunable delays to constrain this "run really fast"
tendency. (Or if you're lucky you can make a special LAN with a tunable
throughput - then you can simulate your target bandwidth more
realisticly).

Cheers,
-- 
Cameron Simpson <cs at zip.com.au> DoD#743
http://www.cskk.ezoshosting.com/cs/

In article 1t8n9hINNq9j at dns1.NMSU.Edu, mcrider at acca.nmsu.edu (Mcrider) writes:
>Could one of you physicist-type cyber-riders give a lucid description/
>explanation of what some folks loosely refer to as a 'tank-slapper'?

An undamped oscillation of camber thrust, with positive feedback,
applied to the front contact patch?  :^)
        - Ed Green, Ed.Green at East.Sun.COM, DoD#0111