[comp.dcom.sys.cisco] SLIP problems using cisco MSM

m33205@is.Morgan.COM (Johnson Lu) (10/09/90)

We have a MSM that we use for dial in SLIP connections, but we don't seem to be
able to get a consistant and satisfactory performance from it.


   +-------+                       +---------+      |               
   | Work- |                       | MSM-32  |      |
   | stn w/|====(dial up @ 9.6k)===| ports;  |***(ethernet)***
   | slip  |                       | 4 runn- |      |
   +-------+                       | ing slip|      |       +-------+
                                   +---------+      |       | Host  |
                                                    |=======| w/    |
                                                            | stuff |
                                                            +-------+ 

Whenever we do a FTP and/or Telent from the Host to the slip worksation, we see
repeated packets and retransmission timeouts (we used a HP4972 for monitoring)
on the host.
What we see is that the window size of the file receiving workstation is set	at 4096 bytes, which the sending host promptly sends out in 4 packets and in
less than 1 second. (The maximum segment size has been set at about 960 bytes)
But the workstation, within this second, only sees the first packet, and sends
out an ack for that packet somethimes later.  The cumulative time between the 
sending host reaching the windown size limit and the workstation sending back 
an ACK is greater than the alloted TCP timeout parameter for the sending host,
so it retransmitts.  This retransmssion problem happens on both FTP and Telnet;
on FTP it happens at the beginning of a file transfer and goes away after  
awhile; telnet on the other hand, is more unpredictable: whenever a user on the Workstation "cat" or "more" a file, we see these retransmissions happen through-out the entire session and with no apparent pattern.

 We guess that these retransmissions happen because the ethernet host pumps out
a window full of data in less time than a terminal server can resend them out
to the slip workstation, so the workstation couldn't possibley send back the ACKfast enough; therefore the hosts times out and these retransmissions take place.

Our questions to all you mighty and brainy TCP/IP gurus out there are:

1) Are we guessing correctly ?  Will "TCP slow start" or increasing the TCP
   timeout parameter help our cause ?  Are there painless ways to implement 
   these schemes ?

2) Would a different Slip router (other than cisco) help ?  Who else makes
   terminal servers ?

3) Are there alternatives to slip for servicing dial in TCP connections ?

4) Will Hulk Hogan ever regain his WWF title again ?  What happened to GLOW ?

Thank you much.

J.Lu  jlu@is.morgan.com

raj@hpindwa.cup.hp.com (Rick Jones) (10/11/90)

By all means get VJ congestion control and avoidance running on all
your systems. Between Slow-start and the improved timers, most of
those retransmissions should dissapear completely. I have personally
seen reductions of 90+% - mileage varies of course...

rick jones

___   _  ___
|__) /_\  |    Richard Anders Jones   | MPE/XL Networking Engineer
| \_/   \_/    Hewlett-Packard  Co.   | raj@hpindio.cup.hp.com
------------------------------------------------------------------------
Being an employee of a Standards Company, all Standard Disclaimers Apply