13 Aug 1986 23:24:15 CDT

Oops! I meant this to go out the first time. Anybody know a good typing

Date: 13 Aug 1986 23:22:07 CDT
Subject: Re: VAX IP/TCP
To: Douglas M. Olson <dolson@ADA20.ISI.EDU>
In-Reply-To: (Message from "Douglas M. Olson <dolson@ADA20.ISI.EDU>" of 18 Jul 1986 13:13:22 PDT)

   The real solution to support the large number of users you talked about
would be to support TCP/IP on your ethernet and use multiple IP gateways.
   PAD is a generic term for the function of turning certain data streams
into X.25 packets for transmission through a packet network, and then
turning the packets back into the data stream on the other side of the
network. The purpose is to make the network transparent to the devices
generating the data streams. The primary use for PADs is to support
synchronous terminals across packet nets. This nutshell description doesn't
cover every aspect of PADs; its just info. The bottom line is PADs are not the
solution you need.
   I recently sent you a note on the uVAX configuration you were considering.
Again, the bottom line of all I said was that I don't think a uVAX can handle
the simultaneous traffic you suggest. I would probably take two to handle
60 connections and even then I think that might be pushing it since every user
will be Telneting into the uVAX and then DecNetting out to their real host in
the Cluster. FTP users will have to transfer files onto the uVAX, then get
them across the ethernet. The same will probably be true of electronic mail
using SMTP; unless there's a package for the uVAX that will forward mail
through the ethernet. If anybody out there in TCP/IP land knows of software
to make the uVAX a cleaner interface, please let me know.
   I have assumed that you are running DECNET in your Cluster. I think you
should consider using TCP/IP and use gateways into the Milnet. However,
I've heard it mentioned that some folks out at JPL are working on what
has been termed a DDN/Decnet gateway. I assume such a device would translate
between the ARM protocols and Decnet and would be transparent. Again, does
anybody else have more info?

Darrel Beach

This archive was generated by hypermail 2.0b3 on Thu Mar 09 2000 - 14:36:35 GMT