Bling-Bling wrote:
> On Mon, 16 May 2005 09:21:58 +1200, Rob J wrote:
>
>
>>>A properly configured server will have a pre-determined maximum number of
>>>concurrent connections configured and will not accept more, and will drop
>>>any further attempts until the number of concurrent connections falls
>>>below that threshold.
>>>
>>>That maximum number should be set according to the capacity of the
>>>server's hardware, and of the bandwidth available to it.
>>>
>>
>>It may be easy for you to say this with your own server with a tiny
>>amount of traffic passing through it.
>>
>>It's a whole different ballgame when an ISP gets what is effectively a
>>DOS attack with a flood of traffic, who knows, 10x or even more of the
>>normal amount of traffic.
>>
>>At the very least there will be obvious network congestion from the
>>traffic into and out of the server.
>>
>>What do you suppose is happening when the server is dropping connections?
>>Could it be that people are finding it difficult to use the server?
>
>
> If under normal conditions the server is operating *that* close to the
> limit of the bandwidth available to it, then one would have thought that
> it was past time for an additional server and/or additional bandwidth.
>
If you limit access to your server at 100% of its resources,
and an attack maxes it out, your ordinary customers are
denied access. It doesn't matter if you do that or you
don't limit it. The only effect is that your customers
cannot get to the server. A server that is maxed out by a
DoS does NOT normally fall over. Though it looks like it
from the customer's end.
I can't talk for all mail server users, but I'd have thought
that it is rare to specifically to tune a mail server to
*restrict* traffic. Usually the problem is to get the
throughput.
Chers,
Cliff
--
Barzoomian the Martian -
http://barzoomian.blogspot.com