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


There is a reach reduction penalty when increasing the line rate at uplink,
using current FP lasers at ONU.
Following the discussions at the EFM, I think that for the providers,
increasing reach seems a more urgent problem than increasing BW.

Lior Khermosh

-----Original Message-----
[]On Behalf Of Gerry
Sent: Wednesday, February 13, 2002 6:29 PM
To: bob.barrett@xxxxxxxxxxxxxxxxxx; carlosal@xxxxxxxxxxxxxxxxxx
Subject: RE: [EFM] [EFM-P2MP] 10G EPONs

Hi Bob,

Thanks for the invite ;)

> 2.5G, 10G P2MP

In my opinion, I don’t think we should work on 2.5/10 Gbps P2MP at this

-- 1Gbps is our voted (>75%)Objective

-- 1Gbps P2MP is the access sweet spot over the next 5-10 years.
Depending on the number of splits, this is in the range of 30Mbps to
1Gbps, on average, per ONU. I know very, very few businesses, let alone
homes, that have this kind of bandwidth today, less than 0.1% (one in a
thousand) is my guess.

-- 1Gbps symmetric is Ethernet speed

-- 1Gbps component pricing will fall rapidly; this is the cost curve to
ride for a standard that will emerge next year.

-- The MPCP protocol is Frame based, using Frame messages (like PAUSE),
and can scale, later, to higher speeds.

-- No service provider has the capacity today to metro-back-haul
multiple 2.5/10G residental and small business access links.

-- And the big one:  I have not heard service providers request 2.5/10G
EPON. By the way, I have heard them always, every time, without
exception, ask for low cost ;)

So I am in favor of 1Gbps symmetrical MPCP P2MP, which I believe what we
are working on.

Best regards,

Gerry Pesavento

> -----Original Message-----
> From:
> [] On
> Behalf Of Bob Barrett
> Sent: Tuesday, February 12, 2002 11:51 PM
> To: carlosal@xxxxxxxxxxxxxxxxxx
> Cc:;;
> Subject: RE: [EFM] [EFM-P2MP] 10G EPONs
> Oh dear, I have opened another can of worms ;-).
> 2.5G is not a 10x speed, so I guess that would go down like a
> lead balloon at 802,3.
> However 2.5G is a current optical specification. GBICs for
> 2.5G exist. I have no idea what the reach and range
> implications are for 2.5G v 1G, this in not my field. Vipul
> should be able to shed some light on this (no pun intended).
> The economics may work in favour of 2.5G. I have no idea.
> I'll leave that issue to the optics people.
> I would think that the (non CSMA/CD) access control mechanism
> that is specified for 802.3 EFM p2mp will scale to 10G.
> Thereafter it would be an economic and market acceptability
> argument. This doesn't mean that I want to stop the 1G EPON
> work, it just means let's make it scalable to 10G and beyond.
> Let's keep this thread open for debate.
> I would invite Gerry Pesavento to respond please.
> Best regards
> Bob
> -----Original Message-----
> From:
> []On Behalf Of
> carlosal@xxxxxxxxxxxxxxxxxx
> Sent: 12 February 2002 18:02
> To: bob.barrett@xxxxxxxxxxxxxxxxxx
> Cc: david@xxxxxxxxxxxxxx;
> vincent.bemmel@xxxxxxxxxxxx
> Subject: [EFM] [EFM-P2MP] 10G EPONs (it was MPCP: Report message)
> Bob,
> > May be  we should be specifying 10G-EPON given that the service
> > providers are  looking at 30x3 channels of HDTV per PON. By
> the time
> > 1G-EPON is standardised it  could be too little too late (again).
> Serious. Why not investigate this? It may look like overkill,
> but we should at least understand the price/performance here.
> Let us see how do it fare:
> - it does not need to be 1 Gbps, or 10 Gbps. Maybe 2.5 Gbps
> is a good compromise for the near future.
> - higher bit rate can be used to counterbalance the relative
> inneficiency of the MPCP protocol. It's a diminishing return
> equation - the faster the system, the less efficient it will
> turn out because of the fixed parameters such as laser on/off
> times. A sweet spot may lie at some point.
> - how do the cost increase with higher speed optics (more
> than 1 Gbps), compared with a more complex (and efficient)
> MPCP implementation?
> Did someone check this at some point over the past few months?
> [Just to remind, once again: traditional Ethernet was half
> duplex, only 40-60% typical efficiency. It was fast enough
> for a long time, and *much* less expensive than any of the
> alternatives. As technology evolved, it was possible to go
> full duplex, and the rest is history.]
> Carlos Ribeiro
> CTBC Telecom