Cell Relay Archive

[skasera@hss.hns.com]

What are the basic advantages that ATM over IP would give?
>From a purely performance point of view, I neither understand cells-in-frames
nor ATM cells in IP datagrams. Why have small cells packed into larger
datagrams. I see it like this: "Since cell size cannot be changed, so be it. but
then what to do to leverage deployment of ATM networks. One solution is lane,
which deprives the end users of native ATM benefits. Other solution as Jon
suggested is ATM over IP. But i do not see any apparent advantage excepting that
prima facie the master-slave relationship between IP-ATM gets reversed.

regards,
sumit kasera






Jonathan Turner <jst@cs.wustl.edu> on 11/24/99 12:21:17 AM

Please respond to Jonathan Turner <jst@cs.wustl.edu>

To:
cc:    (bcc: Sumit Kasera/HSS)

Subject:  CR: Re: A lesson from the ATM/IP wars






Jim Sackman wrote:
>
> <snip>
>
>     The biggest problem that I see with your proposal is that the QOS that
> ATM gives you would be lost with IP.  I would think that a better combining
> of the two would be to set up SVCs based upon RSVP requests.
>
     If RSVP were implemented this might help, although RSVP's
     model of resource reservation is fundamentally flawed,
     because it separates reservation from the routing
     decision.

     I think you may be missing the point of the periodic
     transmission I am proposing. The purpose of this is to
     force all the TCP flows (which dominate internet traffic)
     to reduce their rates to accommodate the tunnels. This
     probably won't be enough to allow you to make strong
     guarantees, but I think it may be possible to get
     sufficiently good performance to allow multimedia
     applications to operate satisfactorily. Better than
     over standard IP networks alone. However, it will
     take some experimentation to verify this.

Jon