[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: 45 thin clients = slow squeeze



Dear Ole-Anders,

thanks for your quite detailed heavy-load-test:

Am Montag, 4. Juni 2012 schrieb Ole-Anders Andreassen:
> 45 clients booted as thin clients. Booting takes several minutes.
> All users opening firefox, openoffice and a pdf file (okular).

This sounds like a nearly standardized test environment.
Maybe, for QA, this could be a start to document how to
put heavy load on a networt - and how to tell the bottle necks.

> The result was a server load of 60.

Was this equally balanced to all 4 cores? Or where there
processes that stuck to cpu0?

If you are in the lucky position to run such a normalized
load-test, you should, please, use the opportunity in order to
check:
- htop: Which processes where the cpu blockers?
- nload: was one of the server's NICs congested?

I assume, you are running a combined tjener/ltsp?
Are there further clients causing traffic at your site?
Is eth1 used for the ltsp network?

What graphic resolution is used on thin clients?
Do you use encrypted traffic for them?

Might Petters "killer daemon" help out here?

Did you try the same test with earlier versions?

Well, I am just a teacher - but eager to know what
other readers of this list will suggest.
Good luck & greetins
Ralf


Reply to: