LSTSRV-L Archives

LISTSERV Site Administrators' Forum

LSTSRV-L

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Topic: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Eric Thomas <[log in to unmask]>
Thu, 9 Feb 1995 23:48:52 +0100
text/plain (53 lines)
FYI -  the "INTERBIT problem"  in question,  for those not  familiar with
BITNET, is the source of most  of the delivery delays we have experienced
since January.
 
  Eric
 
*---------------------- Original message follows -----------------------*
 
CREN has been attacking the INTERBIT problem from three directions.
Briefly they are:
 
1)  Provide Listproc so the list management function can be moved to
smaller, cheaper Unix systems.
 
2)  Provide a high volume mail transfer agent for Unix systems
(ZMailer).
 
3)  Provide P/370s to be installed in RS/6000s so the P/370 can do NJE
routing and the RS/6000 deliver Internet mail.
 
Number 1 is available today.  Some of our members are doing a lot of
mail using Listproc and ZMailer.  I've asked Marco (our Listproc
specialist) to put together some information about how much mail some of
our sites are handling and their hardware configurations.
 
Number 2 is available but not as easy to install and use as we'd like.
Marco has been working with the current maintainer of ZMailer to improve
installation, use and performance and also for a port to AIX so it could
be combined with number 3.  We've also done some experimenting with
ZMailer on a Sun acting as the delivery agent for SMTP on VM and have
seen some very promising results but not yet a full INTERBIT replacement
for a large core site.
 
Number 3 has been the biggest disappointment so far.  It requires IBM do
develop ethernet drivers on the RS/6000 to allow the P/370 to use the
RS/6000 ethernet interface (they already exist on the PS/2).  We've been
hoping for this but haven't heard a thing from IBM about it.  The token
ring drivers do exist so as a short term workaround Arty Ecock has been
working on a solution where the P/370 uses the token ring drivers, the
RS/6000 and SNA to define an SNALINK for TCP/IP.  Obviously this is not
ideal since it is a lot of overhead and requires token ring and SNA but
its a start in the direction.
 
These are CREN's ideas for removing the huge INTERBIT load from IBM
mainframes.  None of these solve the entire INTERBIT problem for sites
doing the highest volume.  Our goal is to provide some relief in the
short term and solve the entire problem in the long term.
 
I wonder if further discussion should take place on the INTERIBT list?
It is open to subscription by anyone.
 
Larry

ATOM RSS1 RSS2