Re: Long distance links
Dan,
I'll intersperse my comments in your note.
"DOVE,DANIEL J (HP-Roseville,ex1)" wrote:
> I can't believe the level of confusion this reflector brings me! :)
>
> I don't interpret Paul's comments to mean that SONET/OC-192 must
> be THE 10GbE PHY, but I am sure that he would like it to be the
> 10GbE WAN PHY which is just fine with me.
I'd certainly like to hear agreement from Paul and the other WAN guys on
your assumption. Personally, I agree with a direction and would support an
objective for a 10 GbE SONET WAN PHY. My interpretation of this objective
is as follows:
1) 10 Gbps MII Interface support: The 10 GbE SONET WAN PHY would use the
same MII interface specified for ALL 10 GbE PHYs. This means the support of
a common MAC/PLS data rate. That rate must meet the data rates of the
highest speed PHY. Since the 10 GbE LAN PHY requires the highest data rate
of 10.0 Gbps, the MAC/PLS data rate is required to be 10.0 Gbps. Throttling
mechanism(s) are required to satisfy the data rate requirements of lower
speed PHYs, if applicable. Note that the MII may be integrated into a
MAC/PHY implementation;
2) 10 Gbps transceiver interface support: It would be extremely useful to
have a common transceiver interface to satisfy the requirements for a
common interface for all 10 GbE LAN PHYs as well as accommodate a new 10
GbE SONET WAN PHY. Howard Frazier, et. al., have proposed a quad 2.5 Gbps
serial differential data interface between a (hopefully pluggable)
transceiver and MAC/PHY logic. It is difficult to envision this interface
being logically integrated into a completely integrated MAC/PHY
implementation except for special purpose implementations;
3) 10 GbE WAN PHY development forum: Based on the intensity and nature of
discussion over this reflector regarding the support of the WAN and the
complete lack of consistent objectives for this effort, I personally view
this effort as being disruptive and detrimental to the timely completion of
an integrated 10 GbE standard. Therefore, I would suggest spinning off the
10 GbE SONET WAN PHY effort as a separate 802.3 activity. This direction is
similar to that taken early in GbE development by spinning off the GbE UTP
PHY as a separate effort. It is also in line with the recent decision by
the HSSG to reject a 10 GbE short-haul copper PHY which has virtually no
conflict with HSSG objectives other than using copper. This effort will
likely proceed with a Call for Interest in September/November.
> However, I still don't understand the persistence that the MAC/PLS
> must run at any speed other than 10.0000Gbps. I think we all realize
> that a 10.0000Gbps interface at the MAC/PLS is going to yield something
> less than a 10.0000Gbps throughput no matter WHAT PHY is running below
> it due to the overheads of preamble, delimiters, and IPG... not to
> mention a possible SONET/OC-192 PHY and it's particular issues.
Yup, I have a very hard time understanding the logic behind requiring a
MAC/PLS rate other than 10.0 Gbps.
> I sure wish we could come to consensus on the fact that it is not
> only possible to accomodate OC-192 with a 10.0000 Gig MAC, but that
> this would provide the most efficient solution from the standpoint
> of product implementation AND standards work.
Agreed.
> It makes sense to put PLLs and FIFOs in PHY components rather than
> to burden the MACs with such things. The additional complexity at a
> WAN PHY is where SONET/OC-192 complexity belongs. Why are we not
> converging upon these points?
Agreed.
> Let's revisit Paul's statements...
>
> > -------------- from Paul --------------------------
> > To carry Ethernet frames on a SONET system:
> >
> > 1)Ethernet must match the payload rate of 9.584640 Gbps
> > 2)Ethernet frames must be encoded with NRZ efficiency
> > 3)Ethernet frame delimiting must operate without
> > special symbols
> > ------------ end from Paul ------------------------
>
> Couldn't this be simply resolved by stating that the WAN
> PHY must;
>
> 1) Resolve the rate mismatch by using HOLD or some other mechanism
> 2) Encode data using NRZ
> 3) Strip Ethernet frame delimiters and re-insert them as necessary
>
> Again, these seem easily accomplished within what is called a PHY.
>
> Look at the complexity of a 1000BASE-T PHY and you should see that
> once the data gets below the MII, it doesn't really matter what
> happens to it as long as it gets to the other port's MII in the form
> of an Ethernet frame.
As another example, I have proposed a 10 GbE PHY on behalf of Transcendata,
Inc., which is integrated within a transceiver/module. The primary reason
for this is that at a line rate of 5 GBaud, all high speed logic must be
contained and located within several mm of the laser/receiver die. Since
most of the logic IS high-speed logic, a chip embodying much of the PHY is
collocated with the optics (could be short-haul copper). To get to the
point, the interface from the two transceivers on either side of the
full-duplex media are quad 2.5 Gbps interfaces. Since this interface is
common for all 10 Gbps PHYs, it is likely to have a common encoding. It's
not likely that the encoding for this interface will be PAM5 as is the
proposal for Transcendata's MAS PHY. This means that data between two MACs
at opposite ends of the link would be framed and delimited with special
characters and then encoded, passed to the transceiver, coding and special
characters would be removed, the data would be recoded as PAM5 symbols,
passed across the media and into the remote transceiver, PAM5 coding would
be removed, transceiver interface coding and special characters would be
applied...
Like Dan, I don't understand what the big deal is about requirements for
Ethernet frames having NRZ coding efficiency and not using special
characters. If these are 10 GbE SONET WAN PHY requirements, then the
proponents of this scheme should go for it. in my mind, it's just another
PHY, it needs to get in line with all other PHY proposals and not impose
its superfluous requirements on all other 10 GbE PHYs.
Best Regards,
Rich
--
> Regards,
>
> Dan Dove
> ..
--------------------------------------------------------------
Richard Taborek Sr. Tel: 650 210 8800 x101 or 408 370 9233
Principal Architect Fax: 650 940 1898 or 408 374 3645
Transcendata, Inc. Email: rtaborek@xxxxxxxxxxxxxxxx
1029 Corporation Way http://www.transcendata.com
Palo Alto, CA 94303-4305 Alt email: rtaborek@xxxxxxxxxxxxx