The IP over ATM Mailing List Archive by date

Cell Relay Retreat>List Archive>month:1994-Sep> msg00174



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

one tcp per vc ???

  • From: Cheh Goh <cng@otter.hpl.hp.com>
  • Date: Wed, 28 Sep 94 17:32:36 BST

> > > 16 million VCs over a 4.8Gbps link means that the average speed of
> > > connections is 300 bps.
> > > 
> > > So, long before you need 16M connections, you need another fiber with
> > > another VC space.
> > > 
> > 
> > Don't be so short sighted.  First of all, not every switch will
> > allow 16 million VC on EVERY port.
> 
> No switch today support such a large number of VCs. But, DON'T BE
> SO SHORT SIGHTED.

I am not saying that it cannot be done.  You are missing the
point here.  I am saying that for the sake of saving cost, 
vendors may not do it for a long time, until users demand it.
Users will scream when VCI runs out.  Why would that happen?  
It will be due to application writers using masses of connections.

I wonder what is the number of VCI per port supported in the switches
available today?  And in the near future?  Granted, if every port 
is going to provide 16 million VCI, there would be no need for 
cause 45: "no VPCI/VCI available".  It exists because of the 
reason I suggested.


Cheh