Re: wanted: wireless magic tricks
Catch is, the bw x delay product will kill you on these links if you run TCP sessions and you won't get anywhere near line rate.
Not true. If you turn on MS TCP or solaris TCP or whatever you won't get line rate out of the box awithout tweaking things a bit, but you can make *TCP* do it. Some folks here have pushed TCP to over 500 Mbps over an OC-12 satellite channel with the normal, everyday TCP implementation in solaris. (I.e., when you discount header overhead, etc. they got very close to line rate). allman --- Mark Allman -- BBN/NASA GRC -- http://roland.grc.nasa.gov/~mallman/
On Thu, Aug 16, 2001 at 01:39:51PM -0400, Mark Allman wrote:
Catch is, the bw x delay product will kill you on these links if you run TCP sessions and you won't get anywhere near line rate.
Not true. If you turn on MS TCP or solaris TCP or whatever you won't get line rate out of the box awithout tweaking things a bit, but you can make *TCP* do it. Some folks here have pushed TCP to over 500 Mbps over an OC-12 satellite channel with the normal, everyday TCP implementation in solaris. (I.e., when you discount header overhead, etc. they got very close to line rate).
Well, that is quite wonderful, but when I approached this problem with a collegue of mine over a sat link for a client that wasn't our experience and after considerable tweaking we ended up having to settle for less. PS: got pointers to documents detailing the 500mbps over OC-12 sat link? email addr will do, as well, I'd love to find out what they did. -- Christian Kuhtz <ck@arch.bellsouth.net> -wk, <ck@gnu.org> -hm Sr. Architect, Engineering & Architecture, BellSouth.net, Atlanta, GA, U.S. "I speak for myself only."
At 13:47 16/08/01 -0400, Christian Kuhtz wrote:
On Thu, Aug 16, 2001 at 01:39:51PM -0400, Mark Allman wrote:
Catch is, the bw x delay product will kill you on these links if you run TCP sessions and you won't get anywhere near line rate.
Not true. If you turn on MS TCP or solaris TCP or whatever you won't get line rate out of the box awithout tweaking things a bit, but you can make *TCP* do it. Some folks here have pushed TCP to over 500 Mbps over an OC-12 satellite channel with the normal, everyday TCP implementation in solaris. (I.e., when you discount header overhead, etc. they got very close to line rate).
Well, that is quite wonderful, but when I approached this problem with a collegue of mine over a sat link for a client that wasn't our experience and after considerable tweaking we ended up having to settle for less.
PS: got pointers to documents detailing the 500mbps over OC-12 sat link? email addr will do, as well, I'd love to find out what they did.
When we ran a T3 satellite link we installed Mentat boxes on each end. See: http://www.mentat.com/skyx/skyx-nasa.html for a nice comparison of TCP vs their SkyX protocol. Bottom line: it works and no user had to change a thing in their TCP stack. Their product goes up to T3 speeds, though. If you have access to the end devices and want to "tune the stack" to achieve full thruput via high delay links see: http://www.psc.edu/networking/perf_tune.html -Hank
-- Christian Kuhtz <ck@arch.bellsouth.net> -wk, <ck@gnu.org> -hm Sr. Architect, Engineering & Architecture, BellSouth.net, Atlanta, GA, U.S. "I speak for myself only."
participants (3)
-
Christian Kuhtz
-
Hank Nussbacher
-
Mark Allman