IETF tweq & ASnet

[Jeff Wabik <[e-mail]>]

I was forwarded this thread by Ross Veach @ uiuc..  Please explicitly
.cc me on any correspondence..  I'll work on getting added to this list.

Some comments, and what I know:

> Date: Mon, 16 Apr 90 18:15:46 EDT
> From: [e-mail] (dave o'leary)

> Does anyone out there know anything about the Army Supercomputer Network?
> We just brought on a new site, the US Army Engineering Water Experimental
> Station (or something like that...).  Currently they are only connected
> to SURAnet, but they have a Milnet gateway coming up soon, and the guy
> there we are working with said they will soon (end of May??) have two 
> more T1's going out as part of the Army Supercomputer Network - one to 
> BRL, up here (only on Milnet, I think...) and another to a site in 
> Michigan, further unspecified.  Anyway, depending on the other sites'
> connections and the routing policy of these circuits it looks like we
> may have Yet Another set of backdoors to worry about.  He said he would
> get me more info about the connections, etc. which I will pass on if 
> it is of interest.

> Date: Mon, 16 Apr 90 21:36:20 -0600
> From: Guy Almes <[e-mail]>

> I was involved in an unsuccessful proposal to secure the Army High
> Performance Computing Research Center, so I know a little of the
> plans, at least as of a year ago.
>   The Army was interested in having several sites connected to each
> other and to the Center.  Among them were the Ballistic Research Labs
> in Aberdeen, the Waterways Experiment Station in Vicksburg, the Tank
> Command in Detroit, and several less important sites in the DC area.
>   The winning proposal came from the Minnesota Supercomputer Center,
> so there is a need to connect this Army network to MSC at high speed.
>   Clearly the traditional MILnet technology is not up to the task, but
> as newer technology is being brought into the DDN, it should be possible
> to meet these needs.
> 
>   From what you say, it sounds like the Army has chosen to build that
> Army Supercomputer Network.  As another geographically national
> discipline/mission-oriented network, it will give us some routing
> opportunities to worry about.
> 
>   A networking person from MSC might be a good person to coordinate
> this and to keep us informed.

I'm the network contact at MSC for the AHPCRC and ASNET, and run
MRNET.   Bob Reschly ([e-mail]) is the contact for the BRL/ASNET
end of things (he should probably be added to this list, too).

TOPOLOGY:



       -------------         -------------
      |   MILNET    |???????|   NSFNet    |
       -------------         -------------
                                  |
       -------------         -------------       -------------
      |   ASNET     |-------|   MRNet     |=====|   MSC       |
       -------------         -------------       -------------


---- is a T1 link
==== is an ethernet attachment.



ASNET connects to the outside world via MRNet (The Minnesota Regional
Network) at T1.  MRNet connects to the NSFNet via NSS12 at T1.  MSC
(Minnesota Supercomputer Center) Connects to MRNet via ethernet.  



ASNET is geographically located in Washington DC at BRL.  MRNet is in
Minneapolis, MN, in the MSC facility.  MSC, obviously, is in
Minneapolis.  The actual ASNET backbone is an ethernet in-a-can
(multiport), with ports currently conencting to MRNet, BRL, WSMR (White
Sands) at 56k, ARDEC (Army Research and Development Command) at 56k,
and HDL (Harry Diamond Labs) at T1.  COE-WES (Corps of Engineers) and
CAA (Concepts Analysis Agency) will be ASNET connected sometime soon.
MRNet is a regional network that provides Internet connectivity
to all of Minnesota.

ASNET sites BRL, WSMR, ARDEC, HDL, and CAA have at least one net on
MILNET, and those are announced to NSFNet by MILNET.  ASNET (138.18) is
connected to NSFNet via MRNet, and should be announced to MILNET via
NSFNet.  ASNET is not directly attached to MILNET.

The plan is to announce several additional nets from each organization
(15 in all) to NSFNet via MRNet, while maintaining their MILNET
connectivity and using MILNET as a backup.  We're working now to
determine the metrics that each net is announced via MILNET with such
that we can provide optimal connectivity from NSFNet.  The ASNet sites
prefer to use the SITE->ASNET->MRNet->NSFNet path for the performance.

This is a reasonable general overview..   If there are additional
questions, fire away.

	-Jeff

-- 
Jeff Wabik                              E/Mail:  [e-mail]
Minnesota Supercomputer Center          AT&T:    [phone]
Minneapolis, MN                         FAX:     [phone]