Cell Relay Archive

[albert.e.manfredi@boeing.com]

In article
<FE7B8F19ED647D12.8C93AEE05CE1F949.33C13CE50488AC7D@lp.airnews.net>,
  Brad Reese <reese@alliancedatacom.com> wrote:
> Hi again Miguel,
>
> Probably the best "Cells In Frames" web page is the 3Com Cells In
Frames
> web page:
>
> http://www.alliancedatacom.com/3com-cells-in-frames.htm
>
> The web page above gives comprehensive information along with
> outstanding visual diagrams.

Very interesting site. The point remains, though, that within the
Ethernet segments you have no QoS. The hope is that the Ethernet load
is light enough that whatever VCs your Ethernet-connected end system
has set up can provide the QoS guarantees that had been negotiated
during the setup phase of each VC from that end system.

The fundamental point is that cells in frames is a concept that allows
a cheap Ethernet interface at the desktop to be used instead of an ATM
NIC (at the desktop). Using switched Ethernet, where each end system
has a dedicated interface, helps, but still cannot be considered a
guarantee of QoS. I mean, you will have to worry about potential for
frame buffering at the switch, which would easily play havoc with any
CDV you might have negotiated for any particular VC. A shared Ethernet
would make the CDV problem more acute. And CLR would be pretty hard to
guarantee.

You can also perhaps implement prioritized queuing within those
Ethernet segments, to help "guarantee" different QoS for each VC set up
by each end system, but now you have to wonder what ATM cells in that
payload buy you.

I think the point that CiF is the opposite of LANE was excellent, by
the way. It pretty much makes the point that both schemes are obviously
compromised. As IEEE 802.3 and IP both add in features to attempt to
implement QoS guarantees, then the idea of sending ATM cells inside the
Ethernet frame sort of loses appeal. The appeal before was only in so
far as the backbone of the network was truly ATM, and the Ethernet
portion was simply a convenient end system interface to the ATM switch.
Lightly loaded, the Ethernet segment can be considered a cheap and fast
physical layer. Heavily loaded, that Ethernet segment will undo what
QoS guarantees ATM tries to provide.

--
Bert
manfredi@arl.bna.boeing.com


Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.