|Thread Links||Date Links|
|Thread Prev||Thread Next||Thread Index||Date Prev||Date Next||Date Index|
To Hal's point, one question I would ask is the following: can "defining a PHY" be interpreted to be inclusive of defining the devices that use that PHY (or at least doesn't constrain you from doing that additional work if it is deemed necessary). If so, then I think we're just fine; if not, then Hal may have a point that at least needs to be considered.
FYI, my hope and expectation is for the former: that it can be interpreted as inclusive, or at least not exclusive.
From: Marek Hajduczenia <marek.hajduczenia@xxxxxxxxx>
Reply-To: Marek Hajduczenia <marek.hajduczenia@xxxxxxxxx>
Date: Mon, 5 Mar 2012 00:14:54 -0700
To: "STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx" <STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx>
Subject: Re: [802.3_EPOC] What is the problem with the EPoC converter, and what is the CLT?
I disagree that the scope needs changing. It does imply two things: (1) we do a new PHY, and (2) it uses EPON protocol. It does not imply or mandate that we need to support full EPON data rates. In fact, that is something that SG and then TF need to work on identifying i.e. whether there are mechanisms which could prevent coax section from being hammered with data. I am quite sure such solutions exist but we need to look at them from architectural perspective at the next meeting and decide what serves SG interest best.
I agree with you and the others that have replied that a bridge is likely necessary to allow flexibility in channel size (and hence CLT capacity). This would also avoid the data granularity problems that strict application of the EPON TDM/TDMA MAC would imply.
WRT, “The THIRD KEY POINT is that I want the CNU to be functionally equivalent to the ONU so that the OLT does not know the difference.”
While this is a good goal, if flexibility in the coax spectrum is desired by the MSOs (channels smaller than 100 MHz) the CNUs will differ from ONUs in at least one respect, they will not have access to the entire EPON capacity. At a minimum it would seem that the OLT scheduler must know the spectral capacity that has been provisioned on the CLT (number and size of channels and the modulation/encoding level the plant will support) and also know which CNUs are associated with that CLT.
On a related subject, the CFI scope is defined as:
"A new PHY for operating the EPON protocol over Coaxial Distribution Networks (“EPoC”)"
Question to the group: In light of these discussions doesn’t this imply the wording of the scope needs to change?
Thanks Valy! Having a bridge in place of what I generically called a converter makes total sense to me. I was going to call it that, and then decided to just stay with the more generic term "converter". But, the device has to bridge because the interface characteristics will be different, not just in modulation characteristics, but in frame structure, data rates, etc. The bridge will receive a frame, convert it to the other frame format, and deliver it to the other media. In the process there will need to me buffering because one will likely be faster than the other (EPON faster than EPoC in most cases).
I totally understand your concern and tried to raise this in the past meeting in Long Beach.
The conversion box in your description is simply called a bridge and is already defined by 802. That bridge has 2 interfaces, EPON for the fiber and an new one for the Coax.
The challenge, as you correctly formulated, is to be able to manage and provide the CNUs form the OLT. I completely agree with this statement and want to see this on the EPOC objectives.
As a side note, there’ve been discussions about “media converters” or “black magic box” between EPON (fiber) and EPOC (coax). Defining those boxes as PHY layer “media converter” is incorrect in the 802 jargon as it implies that every bit of information from one side is converted to the other side. To be more specific, there is no packet filtering in the media converter as the PHY can’t analyze a data packet (that’s done in the MAC).
I propose that we move away from the self-imposed limitation of “media converter” and PHY only protocol and, have a solid discussion on a practical solution that is able to provide a manageable data service between the OLT and the ONU, taking advantage of EPON and the most efficient technologies available for Coax at this time.
We should not impose unnecessary technical and physical limitation on the coax just to have a protocol identical to EPON but we must ensure that whatever we define, can work with and be manageable from existent EPON OLT equipment.
We must also completely define those “black magic boxes”. There is no point in defining an architecture that relays on an element that is not standardized. I propose we use the well defined functionality of Ethernet switches/bridges for that element.
Principal System Architect
Paul, David, Howard, and EPoC Study Group members,
I'm expanding a discussion that I had with Marek, Ed, and Mark to the entire team. I know this will get unruly, but I see this as a "white elephant in the room" for what seems like some sort of philosophical argument, so might as well get it out in the open. Also, I recognize that this was a subject of discussion during the meeting in Newport Beach, but I did not understand it then and thought it might not be important. I see now that it is a key problem that if not resolved will haunt us forever. So, let's see if we can discuss it via Email and see if we can resolve it before the meeting in Hawaii.
My initial statement of the problem to Ed, Mark and Marek, expanded for clarity, is: I struggle with what the CLT is, and what is the problem with the converter that we need to define. I see the EPON and EPoC systems containing these components:
OLT <=== Fiber =====================================> ONUs
OLT <=== Fiber ====> converter <=== HFC network ====> CNUs
The bottom line is that I want to buy a standard OLT, and buy ONUs for customers I can connect via fiber. And, when I can't run fiber to customers, I want to buy a converter between the fiber and the HFC network so I can use the same standard OLT, and use CNUs (an RF version of the ONU) for those customers attached to the HFC network.
The FIRST KEY POINT here is that I want to use the same OLT.
The SECOND KEY POINT is that I want to buy a passthrough device that will be invisible to the OLT, which will take the optical EPON signals and convert them into RF signals. This passthrough device must be flexible in several ways, such as allowing me to use different portions of the RF spectrum, including more and less spectrum as available.
The THIRD KEY POINT is that I want the CNU to be functionally equivalent to the ONU so that the OLT does not know the difference.
I think that I want the RF PHY that the converter and the CNU will use to be defined at IEEE because that should make it easier for the vendors that will implement the converter and the CNU to develop it.
But people tell me that this will be a problem because, from what I understand, the IEEE does not specify converters or some such rationale. Because of that we have to talk about a CLT instead of the OLT, to hide the converter inside the OLT (if I understood correctly).
I hope I was able to keep the definition of the problem simple and clean enough to have a straightforward discussion of why we can't do what I, and my esteemed MSO colleagues, need.
So, what is the problem with the converter, and why is there a need to instead define a CLT which is something I don't want to have?