pace@tabarzin.usace.mil (Joe Pace) (05/30/91)
We have a strange problem when sending data between an Intergraph workstation and an HP 9000/340, the transfer rate seems to be very slow (about 500K/minute) compared to other machines (Sun, Unisys, Apollo) sending the same data to the same HP using the same method. The network looks like this: Sun Apollo Intergraph Unisys | | | | ---+------+-----------+-----------+-----+------------ (Thick net) | PC Router (Ka9q - IBM AT) | -----+----------+----- (Thin Net) | HP 9000/340 The command being issued from the Sun, Apollo, Unisys, and Intergraph is: tar cvfb - 20 . | dd ibs=20b obs=1k | rsh hp-9000 "dd ibs=1k of=data" The file grows at a rate of about 5 megs/minute when using the Sun, Apollo, and Unisys machines, but grows by only about 500K/minute when sending from the Intergraph. The Intergraph machine does not exhibit this problem when sending to other Intergraphs or Apollo machines, and is able to transfer at about 5 megs/minute when sending the data via an Apollo, i.e., ... | rsh apollo "dd bs=1k | rsh hp-9000 'dd ibs=1k of=data' " Using FTP to copy files exhibits the same behavior. Does this sound like a familier problem? It seemd odd that this is only affecting the Intergraoh machines. Thanks Joe -- Joe Pace US Army Corps of Engineers pace@usace.mil Sacramento District JPPACE@UCDAVIS.BITNET 650 Capitol Mall, Sacramento, CA 95814 (916) 551-1133, FAX: 551-1100
raj@hpindwa.cup.hp.com (Rick Jones) (06/03/91)
There are many possibilities when it comes to networking slowness. Obviously, slower systems talk slower. What might not be so obvious is that sometimes faster systems end-up talking slower to slower systems that slower systems do. When this happens you can see a large number of retransmissions on the sending system. The slower system could be the destination, or perhaps an intermediate router. Check rtxs on the Intergraph - particularly if is is equally sow talking with other systems 'on the other side' of the PC router. On newer TCP/IP's, things like 'fast rtx' and VJ congestion control and avoidance can help overcome 'deficiencies' in paths (slow links, routers, etc...) Ask your local Intergraph rep if there is 'fast rtx' or VJ on the system. On such systems, you may still see many rtx's, but the rates will remain higher. If you really want to get your hands dirty, stick an analyser on the wire(s) and trace a few conections. More gory details can be had via email... rick jones ___ _ ___ |__) /_\ | Richard Anders Jones | HP-UX Networking Performance | \_/ \_/ Hewlett-Packard Co. | "It's so fast, that _______" ;-) ------------------------------------------------------------------------ Being an employee of a Standards Company, all Standard Disclaimers Apply
aprm%gd@SHAFTER-EMH2.ARMY.MIL (Gary Dunn) (06/07/91)
Text:
--- Forwarding Mail ----------
Comment:
I passed the posting about problems with an Intergraph to an associate;
here's his reply.
--gd
--- Enclosure ----------------
>From asbp%dmillard Fri Jun 7 14:54:10 1991
Received: by aprm (smail2.5)
id AA03871; 7 Jun 91 14:54:08 HST (Fri)
To: aprm%gd
Subject: Re: Strange slowness
Cc: asbp%dmillard
Date: Fri Jun 7 14:52:17 1991
Message-Id: <9106071454.AA03871@aprm>
From: asbp%dmillard (OpenNET Mail)
Gary
The problem is most likely poor TCP/IP on the I tergraph. Likely
many, many, many retransmissions. The guy should put an analyser on the 802.3
Dave
--- End of Enclosure ---------
--- End of Message -----------