Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!rice!brazos.rice.edu!dboyes
From: dboyes@brazos.rice.edu (David Boyes)
Newsgroups: comp.protocols.ibm
Subject: Re: Using TCP/IP as a gateway to LANs
Message-ID: <1991Jun11.041303.19424@rice.edu>
Date: 11 Jun 91 04:13:03 GMT
References: <1991Jun6.172328.13742@zorch.SF-Bay.ORG>
Sender: news@rice.edu (News)
Organization: Rice University, Houston, Texas
Lines: 43

In article <1991Jun6.172328.13742@zorch.SF-Bay.ORG> jeg@zorch.SF-Bay.ORG (John E. Girard) writes:
>
>The IBM TCP/IP offerings look great in theory,

Works pretty well in practice, too.

>BUT---I hear that the performance of IBM's TCP/IP
>is not soo good.  The 3172 connection from the LAN
>to the Mainframe doesn't do much work and so the
>Mainframe chews cycles managing IP packets.
>I have heard that the maximum throughput is 600Kbps.

Depends greatly on the Ethernet interface you're using. Our
TCP/IP server supports 200+ simultaneous connections w/o using
much more than 3-5% of the available CPU. The 8232 and 3172
aren't as efficient as the Bus Technologies ELC-2 box, but are
more flexible in terms of network media. (ELC-2 does Ethernet
only, 8232 and 3172 can do FDDI and token ring as well as
Ethernet.) We get better throughput than 600Kbps/sec on the
ELC-2s. 


>Is anyone using TPC/IP as an alternative to a
>typical PC gateway, and how do you like it?  How
>is the performance working out for you?  What has
>anyone else heard about the future performance
>of the TCP/IP products?

Works great. This solution is a lot cheaper than using gateway
PCs, although LAN interaction with TCP can be a real problem,
esp. with LAN programs that assume they own the network adapter
exclusively, eg. Netware. Performance is snappy, and it's a lot
more scalable.

>John Girard   New Science Associates, Inc.  jeg@zorch.SF-Bay.ORG
>VOICE: 415-968-3324 ************************** FAX: 415-968-0862


-- 
David Boyes       |The three most dangerous things in the world:
dboyes@rice.edu   |  1) a programmer with a soldering iron,
                  |  2) a hardware type with a program patch, and
"Delays, delays!" |  3) a user with an idea.