obrien@aero.aero.org (Michael O'Brien) (10/10/90)
Someone recently asked a question about use of TCP/IP over a satellite link, and since I've never worked in that realm, the ground got mushy under my feet. I'd like a reality check. I seem to recall that TCP/IP is perfectly usable over a satellite link that has reasonably high bandwidth, but 300msec propagation delays. Adaptive retransmission and multi-packet frames get around this problem, no? Of course, char-at-a-time telnet will lose no matter what, but mail and file transfers should move acceptably, right? More specifically I'm talking about something like a KA9Q implementation running via satellite radio link from a research vessel at sea. -- Mike O'Brien obrien@aerospace.aero.org
ddrg@mentor.gandalf.ca (Duncan Glendinning) (10/10/90)
In article <88084@aerospace.AERO.ORG> obrien@aero.aero.org (Michael O'Brien) writes: >Someone recently asked a question about use of TCP/IP over a satellite >link, and since I've never worked in that realm, the ground got mushy >under my feet. I'd like a reality check. > >I seem to recall that TCP/IP is perfectly usable over a satellite link >that has reasonably high bandwidth, but 300msec propagation delays. >Adaptive retransmission and multi-packet frames get around this problem, >no? Of course, char-at-a-time telnet will lose no matter what, but >mail and file transfers should move acceptably, right? > >More specifically I'm talking about something like a KA9Q implementation >running via satellite radio link from a research vessel at sea. >-- >Mike O'Brien >obrien@aerospace.aero.org We've successfully run SLIP over a satelite link between our offices here in Ottawa, and both Wheeling IL, and Warrington UK. Though a bit slow for interactive use (9600 baud) it was possible. The primary use was for large (>1M byte) file transfers, where we saw better throughput than kermit. Here are some results: Wheeling IL to Ottawa Ont: file size: 2744320 bytes ftp: 76 min kermit: 995 min Please note that the 9600 baud channel is just one of a number that are stat muxed. -- Duncan Glendinning CAnet: ddrg@mentor.gandalf.ca, ddrg@gandalf.ca Gandalf Data Ltd. Voice: (613) 723-6500 Nepean, Ontario Fax: (613) 226-1717 Canada K2E 7M4
mcc@WLV.IMSD.CONTEL.COM (Merton Campbell Crockett) (10/11/90)
It works fine. TELNET on the other hand is something of a pain. I have done a TELNET connection between Stuttgart-Vaihingen, FRG and Westlake Village CA (LA area) with two satellite hops and a 9600 baud link between Westlake and Marina Del Rey. Found you forget what you typed and can almost go out- side for a smoke waiting for the echo. Merton
mleech@bwdlh131.bnr.ca (Marcus Leech) (10/11/90)
In article <88084@aerospace.AERO.ORG>, obrien@aero.aero.org (Michael O'Brien) writes: > More specifically I'm talking about something like a KA9Q implementation > running via satellite radio link from a research vessel at sea. KA9Q works very well in this type of environment--Phils round-trip-timer algorithm is generally acknowledged to be the best in the business. It's quite accurate, and converges quickly. I've used KA9Q over a satellite link between Ottawa, and Calgary. Provided you have a low BER, you can simply crank up your TCP WINDOW and MSS and expect to get quite good performance for FTP, etc. ----------------- Marcus Leech, 4Y11 Bell-Northern Research |opinions expressed mleech@bnr.ca P.O. Box 3511, Stn. C |are my own, and not VE3MDL@VE3JF.ON.CAN.NA Ottawa, ON, CANADA |necessarily BNRs
J.Crowcroft@CS.UCL.AC.UK (Jon Crowcroft) (10/12/90)
>>Someone recently asked a question about use of TCP/IP over a satellite >>link, and since I've never worked in that realm, the ground got mushy >>under my feet. I'd like a reality check. You should see paper by Seo et al in sigcomm 88 for measurement of TCP & IP over SATNET - the relevant bits include: slow start and mean+variance RTT estimator TCP over the atalantic packet satellite network - even a mere RSRE algoritm RTT TCP ran in service over SATNET for a decade ...it just isnt a problem any more :-) jon
mcc@WLV.IMSD.CONTEL.COM (Merton Campbell Crockett) (10/17/90)
David: As the author you well know that a valid TELNET line mode didn't exist until you did the necessary work. 4.3 BSD always devolves to an equivalent of an rlogin connection with single character frames. There are toooo many products for other platforms that don't really support all TELNET options and in fact don't work when faced with a system that doesn't devolve to the 4.3 BSD default. Now that you've completed a legitimate line mode are you going to implement a valid NVDET option? Merton