El 20/10/13 13:03, Antony Stone escribió:
On Sunday 20 October 2013 at 16:50, Alfredo Rezinovsky wrote:
I need to limit the INBOUND bandwith to squid
Why?
I need the http traffic to be between limit.
Using delay pools I can limit the traffic to the clients. I need to
limit the traffic from the internet, leaving at wire speed the traffic
to the clients.
What speed is your Internet link?
What is "wire speed" in your network?
Wire speed is 100Mbps or 1Gbps.
Is that possible?
Realistically, only your ISP can do this.
After all, once the packets have reached your router, you either use them or
you throw them away. If they're part of a TCP stream, throwing them away just
means the other end will re-send them.
There's nothing you can do to stop a packet arriving at your router - you can
only decide what to do with it afterwards.
The package arrives to my router but it can be delayed before reaches
app layer.
You _can_ try to throttle the rate of your outbound acknowledgement packets,
in order to limit the rate at which new packets arrive, but this is:
a) very complicated - you have to throttle each connection from different
sources on the Internet independently
b) very approximate - you can only *try* to limit the rate at which packets
get sent to you (and this may not work how you expect), and there's nothing
you can do about the amount of data in each packet (you can only assume that
each packet is MTU-sized)
c) generally unlikely to give a worthwhile result.
Oh, and by the way, this is not a Squid question. If you want more guidance
on what is possible, you should investigate IPtables and http://www.lartc.org
I know how to do it with qsdisc in linux. I can queue them and let them
pass with delay.
All I want to know is if I can do this with squid.
--
Alfrenovsky
