On Tuesday 14 May 2002 11:05, Alexander Dudko wrote:
> What's your idea of the appropriate TCP window in this case (128
> Kbit per second)?
No more than a couple times the per-user refill rate but not
less than 4KB.
> Understood.
> So, in this case, when, for example, 20 users have their 400 (in
> total 8000) but whole bandwidth is 16000,
> is there any ability to exceed their limits and give them all
> 16000? But at the same time, if all 40 "allowed_hosts" use squid
> then no one takes more than 400.
> In other words, I try to divide all traffic equally between all
> connected clients hosts dynamically. If 10 computers are switched
> on then the bandwidth of each hosts is 16000 / 10 = 1600. But If 40
> computers are switched on then the bandwidth of each hosts is 16000
> / 40 = 400. Can squid-2.4.STABLE6 make it?
> And if yes, could You write example of configuration.
You will have to assign a larger pool size and then rely on Squid to
distribute the shared bandwidth evenly when there is congestion for
the available total bandwidth. The intention is that Squid should
distribute the shared pool pandwidth somewhat even between your
active users, but as always there might be room for improvement.
I would also recommend investigating the effects of using a larger
bucket size, allowing for burstiness. This way you can make sure that
web surfers get a somewhat reasonable bandwidth, while downloaders
are limited. By using a bucket size larger than the refill rate you
can make the limit an average over a longer timeperiod than a single
second at a time...
-- MARA Systems AB, Giving you basic free Squid support Your source of advanced web reverse proxying solutions http://www.marasystems.com/products/Received on Tue May 14 2002 - 03:38:23 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:08:06 MST