Cell Relay Archive

[manfredi@arl.bna.boeing.com]

In article <SAINUL-H.96Dec20183822@eagle.fedu.uec.ac.jp>,
  sainul-h@eagle.fedu.uec.ac.jp (Hossain Sainul) wrote:

> I am studying ATM network.Sometimes I find that ATM does not
> permit cell-retransmission.But when I go through literaures
> on TCP over ATM, I find that retransmission of cells occur if
> any cell belonging to a packet is lost. This is always
> confusing me.Could someone pls explain---
> 1)whether cell-retransmission occurs in ATM
> 2)if the answer is negative,then how the retransmission of
> cells of any TCP-packet is possible in the same architecture
> of ATM?

Different protocol layers have different tasks.

When IP is sent over ATM, ATM is used in much the same way that Ethernet
is used when IP is sent over Ethernet. If you drop an Ethernet packet
which is carrying an IP frame, what happens? The TCP layer (Layer 4)
notices the problem and the IP frames, starting from the one which was
lost, are retransmitted.

If you're carrying IP over ATM, a lost cell causes the TCP layer to go
through the same recovery mechanism.

In fact, if you lose just one cell's worth of an IP frame, you might as
well discard all the other cells of that frame. It's a waste of time to
carefully send all the other cells, right? So some switches do this. It's
a bit of a bastard setup, in that the ATM layer has to understand things
going on at Layers 3 and sort of 4 too, but it is actually quite clever.

So yes, ATM cells are retransmitted, but ATM doesn't really know this.
What's happening is simply that the TCP layer goes back and re-sends some
IP frames. To ATM, it's just a bunch of new cells.

Bert
manfredi@arl.bna.boeing.com

-------------------==== Posted via Deja News ====-----------------------
      http://www.dejanews.com/     Search, Read, Post to Usenet