Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

RE: [EFM] TDM over Ethernet

John / Ron

TDM in frames works very well in a controlled closed environment like a PON
or a p2p link. It requires significantly more gates (for buffers) than TDM
over fiber along side Ethernet using sub-rate multiplexing at the PHY layer
(which runs steady state with minimal buffers). A lot of the complexity
disappears in the sub-rate muxing implementation. Buffers are almost
non-existent. Jitter and latency are non-issues.

TDM in frames also takes away bandwidth from the Ethernet / EFM link which
the sub-rate muxing does not (sub-rate on a carrier of up to 2.5G). Likewise
TDM on a separate lambda is much more expensive (today) than using sub-rate

The number of gates is a mute point, because if you get 200k gates in a part
does it matter if you use 20k of them or 100k of them?

TDM in frames in a metro multi-hop link is a whole different ball game, as
variable latency is a big issue, until the MEF / GMPLS sort out QoS for us.
It would be logical to aggregate circuits and put them on a separate lambda
in the metro. The scope of 802 does include metro area networks, but not the
scope of EFM.

Best regards


> -----Original Message-----
> From:
> []On Behalf Of John Limb
> Sent: 13 December 2001 15:25
> To: 'Ron Jeffries'; stds-802-3-efm@xxxxxxxxxxxxxxxxxx;
> bob.barrett@xxxxxxxxxxxxxxxxxx
> Cc: 'Charles Cook'; kmccammon@xxxxxxxxxxx
> Subject: [EFM] TDM over Ethernet
> Ron,
> 	Providing TDM services (DS1 etc) over ethernet is normally a bit
> challenging due to the delay variation in ethernet - worse at
> lower speeds.
> However, in the PON environment, with good scheduling (at a higher layer)
> using the request/grant loop (that I think we all agree is within
> scope and
> being standardized) we should be able to get ethernet frames
> carrying TDM to
> the OLT with very little delay variation. A small buffer should
> then restore
> perfect TDM synchronicity. Ron, I think that this is what you
> have in mind,
> but Bob, do you agree?
> John Limb
> -----Original Message-----
> From:
> []On Behalf Of Ron
> Jeffries
> Sent: Thursday, December 13, 2001 12:05 AM
> To:
> Cc: Charles Cook; kmccammon@xxxxxxxxxxx; btolley@xxxxxxxxx;
> stds-802-3-efm@xxxxxxxxxxxxxxxxxx
> Subject: Re: [EFM] TDM circuit emulation
> Providing "first class citizen" support for TDM over Ethernet
> is both technically sound and something I'd expect some service
> providers will demand. But TDM support over Ethernet doesn't
> require changes to the PHY, or the use of another lambda.
> --ron jeffries
>   rjeffries@xxxxxxxxxxxxxxxxx