Re: PMD Compromise
In conversations that I have had with "carrier" people (more than just my company) the perceptions were that 1500nm lasers would be
tight tolerance and on the ITU grid. I do not believe that standardizing on a high power, tight tolerance lasers at one of the ITU
grid frequencies is what the P802.3ae TF had in mind.
There was also the perception that the data switches would replace the LTE network elements. This is a mistake that we (the service
providers) made with POS. I was part of that at the time and will readily admit that I, and many others were wrong. A normal,
customer data switch is not, can never be, and should never be perceived as an active carrier network element. There are carrier
network element systems that are in the works that will have 802.3 switching functionality in them, that are designed specifically
to be network elements, using full SONET (EoS). One vendor already has one on the market. The P802.3ae TF has not assumed the
charter to standardize on this type of equipment.
True, 1300nm does not lend itself to EDFA amplifiers. By the time that someone is paying for amplifiers, they are operating at a
cost factor way out of the range of what I believe the TF invisions. As such, amplifiers will be owned by the service or fiber
providers not the enterprise customer. The service provider will want a transducer between the customer and his fiber plant in
order to support any SLAs that are in place. The transducer will move the 1300nm Ethernet WAN signal to the 1500nm, ITU grid space.
1500nm lasers are not necessary for the Ethernet WAN PHY to operate in the very long reach systems. This is true of a lot of legacy
OC48 transmission equipment today. The transducer/amplifier scenario is exactly what happens today in the case of the older
transmission network elements that have 1300nm lasers.
The miss-perception on the part of service providers is natural when you consider that 1500nm lasers are more in the domain of
"carrier" space than anywhere else. By not having 1500nm in the specification, the perception would be that the systems defined by
the P802.3ae specification are not carrier type network elements. If a vendor wants to enter the carrier network element market
space with an P802.3ae data switch, that vendor would then fall under T1X1 specifications, not IEEE.
I think that we need to make a clear distinction between the T1X1 based EoS proposed standard and the P802.3ae proposed standard.
One of the simplest ways to do that is by making the 40km specification at 1300nm, not 1500nm. If system vendors want to provide
1500nm lasers at power specifications beyond 40km, then they are free to do that outside of the standard, just as several vendors do
today with GbE.
----- Original Message -----
From: "Vipul Bhatt" <firstname.lastname@example.org>
To: "P802.3ae Task Force Reflector" <email@example.com>
Sent: Thursday, June 08, 2000 7:52 PM
Subject: RE: PMD Compromise
> Jay Hoge wrote:
> > Upon reflection, I agree with you on the 1.3u vs.
> > 1.55u smf issue. If people want to go truly long
> > haul in native Ethernet, there needs are
> > outside the scope of this standard.
> No, I think that would be a mistake! This standard can't afford to
> ignore a sizeable market need - how useful is a standard that fails
> to specify a version many people want? Please allow me to argue in
> favor of keeping 1550 nm link on the table.
> Ethernet backbone over dark fiber is appealing. It is a fast and
> affordable alternative to SONET, enabling cost-effective high
> bandwidth connectivity to data centers and Internet. Market is
> changing; so should change the scope of a new standard. Metro links
> are increasingly in demand, and we can't be sure that >40 kms will
> be an insignificant piece of that market. If anything, we are
> getting the opposite message from the market - sales of 80-km GBICs
> are increasing at a rapid rate. We can't afford to ignore these
> trends when drafting a new standard.
> The 1550 nm PMD will be expensive by LAN standards, but it will be a
> bargain compared to SONET monthly rates. As volumes build up, prices
> will drop further.
> In the near future, when we consider 40G operation, a 4:1 WDM of 10G
> Serial links will be one of the major options. Compatibility with
> 1550 nm WDM devices and amplifiers will be required then. I know we
> are not obliged to take this future event into account, but why not
> think about it?
> By refusing to standardize a significant market need, we will be
> depriving the market of some key lubricants - interoperability,
> competition and lower cost. The market will prevail despite us, but
> magazine articles for years to come will keep reminding us of what a
> poor decision we made.