src@scuzzy.mbx.sub.org (Heiko Blume) (08/23/90)
can someone tell me what throughput i can expect with sl/ip over a modem connection with HST modems (14400bps)? we talk to the modems at 19200 and use hardware flow control (FAS with 16550 chips). we did a ftp and the throughput was ~0.2 KB/s on a 600KB file, which is unacceptable. did we do something wrong? (of course we did sldialup/attach 19200 ttyF01). -- Heiko Blume c/o Diakite blume@scuzzy.mbx.sub.org FAX (+49 30) 882 50 65 Kottbusser Damm 28 blume@netmbx.UUCP VOICE (+49 30) 691 88 93 D-1000 Berlin 61 blume@netmbx.de TELEX 184174 intro d scuzzy Any ACU,e 19200 6919520 ogin:--ogin: nuucp ssword: nuucp
larry@nstar.uucp (Larry Snyder) (08/28/90)
src@scuzzy.mbx.sub.org (Heiko Blume) writes: >can someone tell me what throughput i can expect with sl/ip over >a modem connection with HST modems (14400bps)? we talk to the modems >at 19200 and use hardware flow control (FAS with 16550 chips). we did >a ftp and the throughput was ~0.2 KB/s on a 600KB file, which is unacceptable. >did we do something wrong? (of course we did sldialup/attach 19200 ttyF01). SlIP requires the use of duplex modems for best throughput, and the HST is limited by the baud rate of the "back channel" -- Larry Snyder, Northern Star Communications, Notre Dame, IN USA uucp: iuvax!ndmath!nstar!larry -or- larry@nstar Public Access Unix Site (219) 289-0282 (5 lines/PEP/HST/Hayes-V)
cpcahil@virtech.uucp (Conor P. Cahill) (08/29/90)
In article <1990Aug23.160013.1199@scuzzy.mbx.sub.org> src@scuzzy.mbx.sub.org (Heiko Blume) writes: > >can someone tell me what throughput i can expect with sl/ip over >a modem connection with HST modems (14400bps)? we talk to the modems >at 19200 and use hardware flow control (FAS with 16550 chips). we did >a ftp and the throughput was ~0.2 KB/s on a 600KB file, which is unacceptable. >did we do something wrong? (of course we did sldialup/attach 19200 ttyF01). Your problem is probably caused by the fact that the HST modems running at 14.4kbs are not truely full duplex. They (in a similar vain to the t2500) have a high speed channel in one direction and a low speed reverse channel (switching the directions as necessary). This is a really bad scenario for packetized protocols that require acknowledgment packets (unless the modem "spoofs" the protocol, like what the Telebits do to UUCP). Your best bet would be to get a v.32 modem (if you require the HST 14.4 compatibility for other purposes, use an HST dual-standard). Full duplex 9600 baud modems (i.e. v.32) should get around 800 bytes/sec throughput on clear lines. -- Conor P. Cahill (703)430-9247 Virtual Technologies, Inc., uunet!virtech!cpcahil 46030 Manekin Plaza, Suite 160 Sterling, VA 22170
ewv@craycos.com (Eric Varsanyi) (08/30/90)
In article <1990Aug23.160013.1199@scuzzy.mbx.sub.org> src@scuzzy.mbx.sub.org (Heiko Blume) writes: >can someone tell me what throughput i can expect with sl/ip over >a modem connection with HST modems (14400bps)? This does not directly relate to HST's, but it might be interesting anyway. Configuration: -------------- SCO Unix (3.2.2) with the vanilla driver on a vanilla serial board Connection between 2 Telebit T2500's. Cisco terminal server providing slip bridge onto ethernet Transfer of 100 Kb file (binary, Unix kernel from a Sparc) 19200 hardware flow control on both sides Numbers: -------- Kb/s (reported by ftp) Roundtrip (reported by ping) V.32 mode .795 370 (s50=6) Pep .799 1420 Pep/compressed .823 1551 (s110=1) Watching the lights on the modem it looked like Pep's reverse channel was not wide enough to avoid turning the line around every few packets to send back the ACK's. Interactive was virtually useless at 1.4 seconds/echo, but using V.32 even editing with vi is tolerable. Compressed mode isn't worth the trouble unless you are sending BIG files. Does anyone know when/if Telebit is going to put IP header prediction into the modems (so they don't have to turn around for the ACK's)? -Eric Varsanyi Cray Computer Corporation ewv@craycos.com -- -Eric Varsanyi (ewv@craycos.com) Cray Computer Corporation
bob@MorningStar.Com (Bob Sutterfield) (09/01/90)
In article <1990Aug29.233628.2595@craycos.com> ewv@craycos.com (Eric Varsanyi) writes:
SCO Unix (3.2.2) with the vanilla driver on a vanilla serial board
SunOS 4.0.3 on a 4/60 and a 4/110, on console serial ports
Connection between 2 Telebit T2500's.
Connection between a TB+ and a TB2500, using PEP
19200 hardware flow control on both sides
Same here
Kb/s (reported by ftp)
Pep .799
Pep/compressed .823
(s110=1)
We saw those sorts of numbers with vanilla SLIP. Things improved to
around 1.0Kb/s with header-compressed SLIP. We saw numbers like
1.3-1.5Kb/s FTP throughput when using PPP.
Does anyone know when/if Telebit is going to put IP header
prediction into the modems (so they don't have to turn around for
the ACK's)?
Try using a smarter protocol, and the modems will do just fine.
Header compression helps a *lot*.vjs@calcite.UUCP (Vernon Schryver) (09/02/90)
> SlIP requires the use of duplex modems for best throughput, and the HST > is limited by the baud rate of the "back channel" It has been widely reported that you can get some less than 960 Bytes/sec through FTP over SLIP over a v.32 line (e.g. any of several v.32 modems or a Telebit T2500), and somewhat more than 1,100 Bytes/sec through FTP over SLIP over Telebit PEP (e.g. Telebit T2500 or TB+). Among other places, I have heard these reports in the BARRNet community from developers running 386 clones. 9600 b/s V.32 is prefered for most SLIP uses, because the 40 byte TCP/IP headers are painful on half-duplex modem protocols like PEP (there is no reverse channel in PEP). You can (since I have) compress and predict the TCP/IP headers to less than a PEP micro packet to make SLIP not significantly more painful over PEP than cu. In sum, if you are transfering files, use PEP. If you are typing to rlogin, use v.32. (This may not apply if you have 14.4 bps V.32.) Van Jacobson header compression code has been published in an RFC. I'm told it is also in 4.3BSD Reno. Vernon Schryver, vjs@calcite.uucp