On 26/08/09 07:17 PM, Dan McDonald wrote:
On Thu, Aug 27, 2009 at 10:12:33AM +0800, Kacheong Poon wrote:
<SNIP!>
So at 10Mbit, that:
1250000 bytes * 0.185sec == 231250 bytes.
I think the default should be changed.
Perhaps to 256k then? (It would cover your case that way.)
I guess that should be fine for most usage.
Cool! I *thought* there was an RFE already filed. Perhaps not.
Does anyone else in the peanut gallery have any thoughts?
Peanut gallery? :)
So there are two considerations:
1) buffer allocation on sockets when we're running 1000s
of TCP connections: @1M, 10,000 sockets = 20GB of RAM,
@48k, 10,000 sockets = 900MB of RAM
... yes, the buffers aren't allocated "straight away"
but it is a measure of the obligation being advertised.
2) reduced security through a larger window making it easier
to guess sequence numbers. 48k = 1 in ~87,000,
1M = 1 in 4096
Applications that want a bigger window can issue a setsockopt
to expand the kernel buffer size (and thus the window.)
Darren
_______________________________________________
networking-discuss mailing list
[email protected]
