----- Original Message ----- From: "Masood Ahmad Shah" <[EMAIL PROTECTED]> To: "fooler" <[EMAIL PROTECTED]>; "RAHUL T. KARTHA" <[EMAIL PROTECTED]>; <[EMAIL PROTECTED]> Sent: Tuesday, September 02, 2003 7:29 PM Subject: Re: [squid-users] tweeking squid for vsats
> I want to bring your notice that setting you are going to tell Mr Rahul are > by default set into Linux like show below.. > > 1) be sure that the followings are enabled in your /etc/sysctl.conf: > > net.ipv4.tcp_window_scaling=1 > net.ipv4.tcp_timestamps=1 > net.ipv4.tcp_sack=1 > > so no need to change these setting.... mr. masood shah, read my statement in number 1 again.. i said "be sure" that the followings are enabled... im implying that to double check those settings in case he modifies it... > better to check again your upstream > and downstream packet delay...first like i said to my first post, latency is caused by the speed of light... you cant decreased it unless albert einstein is wrong... satellite links have longer path... therefore the longer path you have the higher latency you can get... in general the time to send a packet depends on two factors: 1) propagation delay (caused by the finite speed of light, latencies in transmission equipment, etc) and 2) transmission delay (how many bits per second a media can transmit) tcp window scaling, timestamps and selective acknowledgment are intentionaly designed for huge bandwidth with higher latency links... in order to fully saturate that kind of link, you have to increase your tcp window size using *bandwidth-delay product*... this is one of the options to improve your throughput over satellite links aside from a special protocol which specifically design for satellite links that handled by the satellite providers for higher throughput with lower packet loss... this special protocol encapsulates the tcp/ip packet... fooler.
