scott@csis.oz (Scott Milton) (06/28/89)
Do Trailblazers work well with SLIP? What modems should I use for 9600 Baud SLIP? I have been told that because TCP sends acknowledgments for every packet, that Trailblazers don't work well, as they are designed for large traffic flows in a single direction for longer periods. With SLIP they spend alot of the time working out that they have to change direction. There is presumably also some overhead in doing error correction, which should be handled by the TCP layer? I believe I have missed some discussion of this topic in this group. Apologies. We do not archive news groups. Any suggestions or opinions greatly appreciated. -- ACSnet: scott@csis.oz Scott MILTON (ph +61-62-750923) email: scott@csis.oz.au (fax +61-62-571052) CSIRO Division of Information Technology, Acton, Canberra. ACT. AUSTRALIA.
grr@cbmvax.UUCP (George Robbins) (07/02/89)
In article <219@csis.oz> scott@csis.oz (Scott Milton) writes: > > Do Trailblazers work well with SLIP? > What modems should I use for 9600 Baud SLIP? No, Trailblazers aren't very good for slip, at last until Telebit get of its *ss and delivers roms that understand the slip protocol. The best bet for right now is a V.32 modem, for example the Telebit 2500 or any of the others. You might get good performance with one of the asymetical channel modems like the USR 9600 buad modem, if most of your traffic is in one direction, perhaps someone can testifty about such. -- George Robbins - now working for, uucp: {uunet|pyramid|rutgers}!cbmvax!grr but no way officially representing arpa: cbmvax!grr@uunet.uu.net Commodore, Engineering Department fone: 215-431-9255 (only by moonlite)
henry@utzoo.uucp (Henry Spencer) (07/02/89)
In article <219@csis.oz> scott@csis.oz (Scott Milton) writes: >Do Trailblazers work well with SLIP? >What modems should I use for 9600 Baud SLIP? > >I have been told that because TCP sends acknowledgments for every packet, >that Trailblazers don't work well, as they are designed for large traffic >flows in a single direction for longer periods. With SLIP they spend alot of >the time working out that they have to change direction... Trailblazers in PEP mode are not great for SLIP. With header compression and other tweaks they can be passable; the terminal room at the San Diego Usenix had its Internet connection done that way. It wasn't great, but it did work okay when I tried it -- although people who used it at busier times were less happy, I gather. (Unfortunately the compression code, from Van Jacobson, is not widely available yet.) At present, you're better off with a V.32 modem for 9600-baud SLIP. The Baltimore Usenix terminal room used a Telebit T2500 in V.32 mode, and was a noticeable improvement on San Diego. Other V.32 modems should also be okay, but make sure you get real V.32 (there is at least one half-duplex modem using a protocol related to V.32). Don't expect miracles; try before you buy. There are things that can be done to optimize TCP/IP for slow lines, but your system almost certainly doesn't do them. -- $10 million equals 18 PM | Henry Spencer at U of Toronto Zoology (Pentagon-Minutes). -Tom Neff | uunet!attcan!utzoo!henry henry@zoo.toronto.edu
vjs@rhyolite.wpd.sgi.com (Vernon Schryver) (07/02/89)
In article <219@csis.oz>, scott@csis.oz (Scott Milton) writes: > > Do Trailblazers work well with SLIP? > What modems should I use for 9600 Baud SLIP? > -- > ACSnet: scott@csis.oz Scott MILTON (ph +61-62-750923) > email: scott@csis.oz.au (fax +61-62-571052) > CSIRO Division of Information Technology, Acton, Canberra. ACT. AUSTRALIA. TB+'s work for FTP, RCP, &co over SLIP if you do not compress TCP/IP headers. If you compress headers, they do almost as will as with cu/tip for interactive stuff. If you do not compress TCP/IP headers, TB+'s are not useful with SLIP--at least that's what I've found. The only reason I know to use SLIP is if you want to multiplex a stuff over a single line. Vernon Schryver Silicon Graphics vjs@sgi.com
david@ms.uky.edu (David Herron -- One of the vertebrae) (07/02/89)
Trailblazers, in PEP mode, are good for applications that are
largely unidirectional. Like this interactive news reading
that I'm doing right now. Or a file transfer protocol that
only works in one direction, and doesn't have much back traffic
in ACKing. Like, kermit with long packets should work really well.
SLIP, depending on what you're doing, isn't necessarily one-direction.
Generally speaking it's equal in each direction.
BTW ... I was *really* impressed with the terminal room at Usenix.
Someone was wanting Telebit to get off their *ss and implement SLIP
spoofing. First off, I don't think it would work since PEP doesn't
allow equal data flow in each direction -- yet. But in any case,
I don't think Telebit wants to implement SLIP support until the RFC
is finished...
--
<- David Herron; an MMDF guy <david@ms.uky.edu>
<- ska: David le casse\*' {rutgers,uunet}!ukma!david, david@UKMA.BITNET
<-
<- New word for the day: Obnoxity -- an act of obnoxiousnesswlm@archet.UUCP (William L. Moran Jr.) (07/03/89)
In article <7198@cbmvax.UUCP> grr@cbmvax.UUCP (George Robbins) writes: >In article <219@csis.oz> scott@csis.oz (Scott Milton) writes: >> >> Do Trailblazers work well with SLIP? >> What modems should I use for 9600 Baud SLIP? > >No, Trailblazers aren't very good for slip, at last until Telebit get >of its *ss and delivers roms that understand the slip protocol. > >The best bet for right now is a V.32 modem, for example the Telebit 2500 >or any of the others. You might get good performance with one of the >asymetical channel modems like the USR 9600 buad modem, if most of your >traffic is in one direction, perhaps someone can testifty about such. >-- Another person and I recently set up a slip network. After evaluating modems we settled on telebit t2500s. The v32 performance is excellent, and slip works well with them. Surprisingly, slip works well in pep mode between 2 2500s also. The interactive feel for telnet and rlogin isn't bad, and the ftp performance kicks *ss. V32 mode really wins when you have more than one thing going on over the line at a time. Bill Moran PS I have no connection with telebit other than as a customer ... etc... -- arpa: moran-william@cs.yale.edu or wlm@ibm.com uucp: uunet!bywater!acheron!archet!wlm or decvax!yale!moran-william ------------------------------------------------------------------------------- Idealism is fine, but as it approaches reality the cost becomes prohibitive. wfb
jparnas@larouch.UUCP (Jacob Parnas) (07/03/89)
In article <12036@s.ms.uky.edu> david@ms.uky.edu (David Herron -- One of the vertebrae) writes: >... >SLIP, depending on what you're doing, isn't necessarily one-direction. >Generally speaking it's equal in each direction. >... SLIP is usually going full speed in one direction and without header compression sending several ~ 40 byte packets per second in the other direction. The trailblazer (PEP protocol) does well for with applications that mind high latentcy (like ftp, which gets about 1.1Kbytes/second vs. about .9Kbytes/second using V.32 modems.) but does very poorly on interactive applications like rlogin, telnet, ping, etc without header compression. For instance, ping takes about 1200-1600 ms. in PEP mode vs. about 270ms. with V.32 modems. In a telnet or rlogin, PEP gets 20 characters behind (what you are typing) sometimes, while V.32 gets at most 1 or 2 characters behind). The Telebit T2500 has both PEP and V.32. I get mail and news in PEP mode and use V.32 for SLIP. V.32 wins big over PEP if you are running two applications that want full bandwidth in both directions. For instance, once I got .8 Kilobytes/second on simultaneous ftps in opposite directions in V.32 mode! This has useful benefits. For instance, it is reasonable to rdist from your work machine to your home machine at the same time as you are running rlogin from your work machine to your home machine in V.32 mode, but I think performance would be much worse in PEP mode. ------------------------------------------------------------------------------ | Jacob M. Parnas | DISCLAIMER: The above message is from | | IBM Thomas J. Watson Research Ctr. | me and is not from my employer. IBM | | Arpanet: jparnas@ibm.com | might completely disagree with me. | | Bitnet: jparnas@yktvmx.bitnet \---------------------------------------| | Home: ..!uunet!bywater!acheron!larouch!jparnas | Phone: (914) 945-1635 | ------------------------------------------------------------------------------
cander@unisoft.UUCP (Charles Anderson) (07/04/89)
From article <219@csis.oz>, by scott@csis.oz (Scott Milton): > > Do Trailblazers work well with SLIP? > What modems should I use for 9600 Baud SLIP? > > I have been told that because TCP sends acknowledgments for every packet, > that Trailblazers don't work well, as they are designed for large traffic > flows in a single direction for longer periods. With SLIP they spend alot of > the time working out that they have to change direction. There is presumably > also some overhead in doing error correction, which should be handled by the > TCP layer? I recently set up SLIP on Apollos (a non-trivial task) using Trailblazers running at 9600 baud. I found that SLIP between two serial ports without modems ran at roughly 75% of the maximum throughput (9600 baud) and that with the TB's in the loop performance was roughly 50% of 9600 baud, not too bad of a performance hit. The test I performed was a simple rcp from one node to an other. If I were running rcp's in both directions, the throughput might have been less. Also it should be noted that intereactive performance (using rlogin) over the SLIP line with TB's was real dismal. To say it was "bursty", would be putting it kindly. -- Charles. {sun, amdahl, ucbvax, pyramid, uunet}!unisoft!cander
matt@oddjob.uchicago.edu (Matt Crawford) (07/04/89)
OK, how about the Microcom QX/V.32c? Has anyone run SLIP on that? If so, how high did you crank up the baud rate on a Sun CPU serial port? If you used the data compression feature, how did you set up the flow control on the sun? Some people here have bought the QXs for one purpose and are getting 30 kb/s or more effective throughput over dialup phone lines between Macintoshes. I want to try them out with SLIP between Suns and would like to hear experiences, if there are any. ________________________________________________________ Matt Crawford matt@oddjob.uchicago.edu
vjs@rhyolite.wpd.sgi.com (Vernon Schryver) (07/04/89)
In article <2109@unisoft.UUCP>, cander@unisoft.UUCP (Charles Anderson) writes: > > ...I found that SLIP between two > serial ports without modems ran at roughly 75% of the maximum > throughput (9600 baud) and that with the TB's in the loop performance > was roughly 50% of 9600 baud... > > Charles. > {sun, amdahl, ucbvax, pyramid, uunet}!unisoft!cander I've noticed similarly low numbers for rcp of .5KB/sec between IRIS's over interfaces set to 19.2, but with FTP, I've seen 1.4KByte/sec on the same phone calls. Notice that 1.4KB is about what you can hope for with UUCP over TB's. No TCP retransmissions were glaringly obvious. TB compression does not seem to make much difference. The modem lights look about the same in all cases--almost solid one-way traffic. Given the large buffers of rcp and the large TCP windows on IRIS's, I don't know what's going on. Does anyone else? Vernon Schryver Silicon Graphics vjs@sgi.com
smb@ulysses.homer.nj.att.com (Steven M. Bellovin) (07/05/89)
In literature handed out at Usenix, Trailblazer said that they had header prediection/compression support for IP. It hasn't been released yet, but they stated their intention to support all versions of dial-up IP.
jpr@dasys1.UUCP (Jean-Pierre Radley) (07/11/89)
Am I wrong in suggesting that discussions of Telebit modems, their register settings, and so forth, might get better results in biz.comp.telebit? I think the manufacturer is active in that newsgroup. -- Jean-Pierre Radley CIS: 72160,1341 jpr@jpradley.UUCP
steve@polyslo.CalPoly.EDU (Steve DeJarnett) (07/12/89)
In article <10170@dasys1.UUCP> jpr@dasys1.UUCP (Jean-Pierre Radley) writes: >Am I wrong in suggesting that discussions of Telebit modems, their >register settings, and so forth, might get better results in >biz.comp.telebit? Most of the USENET 'world' does not receive the biz.* groups, and some of us may be precluded from receiving them due to their commercial nature. >I think the manufacturer is active in that newsgroup. They've been active in this group in the past also. ------------------------------------------------------------------------------- | Steve DeJarnett | Smart Mailers -> steve@polyslo.CalPoly.EDU | | Computer Systems Lab | Dumb Mailers -> ..!ucbvax!voder!polyslo!steve | | Cal Poly State Univ. |------------------------------------------------| | San Luis Obispo, CA 93407 | BITNET = Because Idiots Type NETwork | -------------------------------------------------------------------------------
news@investor.UUCP ( Bob Peirce) (07/13/89)
In article <10170@dasys1.UUCP> jpr@dasys1.UUCP (Jean-Pierre Radley) writes: >Am I wrong in suggesting that discussions of Telebit modems, their >register settings, and so forth, might get better results in >biz.comp.telebit? > A lot of usenet doesn't get biz.comp.telebit, whatever that is! I would like to continue to see it here. > >I think the manufacturer is active in that newsgroup. > Why don't they get active here? -- Bob Peirce, Pittsburgh, PA 412-471-5320 uucp: ...!{allegra, bellcore, cadre, idis, psuvax1}!pitt!investor!rbp NOTE: Mail must be < 30K bytes/message