daveb@llama.rtech.UUCP (Dave Brower) (12/08/87)
When I run uucp/kermit at 9600 baud between my 7300 and another machine,
about the best throughput I can get is around 500 chars/sec, or ~4800
effective baud at 9 bits/char. Curiously, if I have two channels
running between the two machines (using a combo card), _each_ will run
500 c/s.
Has anyone got any ideas on how to improve this? What limit am I
running into? Since I do backups from the 7300 to the other machine,
and it take about 20 hours to do one with this throughput, even marginal
gains would help.
(If necessary I can write my own transfer software using VMIN/VTIME
stuff. Is it worth the effort?)
Thanks,
-dB
"I don't care what you say, as long as you spell my name right."
{amdahl, cbosgd, mtxinu, ptsfa, sun}!rtech!daveb daveb@rtech.uucpdca@toylnd.UUCP (David C. Albrecht) (12/17/87)
> Has anyone got any ideas on how to improve this? What limit am I > running into? Since I do backups from the 7300 to the other machine, > and it take about 20 hours to do one with this throughput, even marginal > gains would help. I can't really help you on the throughput. It is possible you could take one of the zmodems that has been posted in the past and it would have better results. If you aren't already, however, I would highly suggest that you take the sources for arc or zoo and get them running on both ends of the pipe. 'arc'ing or 'zoo'ing the data before you transmit it can save you large amounts of time and also make it possible to use transmission protocols which don't support wildcarding. You can leave the transmitted lump in the arc or zoo and it will take up less space on the back-up system as well. David Albrecht
daveb@llama.rtech.UUCP (Dave Brower) (12/22/87)
In article <191@toylnd.UUCP> dca@toylnd.UUCP (David C. Albrecht) writes: >> Has anyone got any ideas on how to improve this? What limit am I >> running into? Since I do backups from the 7300 to the other machine, >> and it take about 20 hours to do one with this throughput, even marginal >> gains would help. > >I can't really help you on the throughput. It is possible you could take >one of the zmodems that has been posted in the past and it would have better >results. > >If you aren't already, however, >I would highly suggest that you take the sources for arc or zoo and get >them running on both ends of the pipe. 'arc'ing or 'zoo'ing the data >before you transmit it can save you large amounts of time and also make >it possible to use transmission protocols which don't support wildcarding. >You can leave the transmitted lump in the arc or zoo and it will take >up less space on the back-up system as well. I've already been compressing the input stream. What _has_ helped is bumping the speed to 19200, which seems to be giving about half again as much throughput as 9600. However, there are side effects: if a uucico decides to call outgoing while a transfer is going on using combo-card ports, the combo-card port can get messed up. And occasionally the whole 7300 will wedge. This is running 3.50. Can someone correct me? I am of the impression that serial port problems were among the reasons for 3.51, but that they weren't really that much improved. I am really in no mood to buy a new set of software to fix a kernel bug. Thanks, -dB "I don't care what you say, as long as you spell my name right." {amdahl, cbosgd, mtxinu, ptsfa, sun}!rtech!daveb daveb@rtech.uucp
davidsen@steinmetz.steinmetz.UUCP (William E. Davidsen Jr) (12/23/87)
You really do want to use zmodem for file transfers whenever possible.
It will run the line at about 91% rated speed (slowing under errors, of
course). It is *very* reliable and features restart so you don't resend
if a connection get totally broken. I had it running on my 7300 bbs and
it worked fine. I also has a hacked up mail and file xfer system that
used it to selected other systems, but all that stuff was on the system
when I sold it.
I can still dig up a copy of the zmodem stuff, although it's been posted
many (many-many) times.
--
bill davidsen (wedu@ge-crd.arpa)
{uunet | philabs | seismo}!steinmetz!crdos1!davidsen
"Stupidity, like virtue, is its own reward" -me