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

Re: What is ~10 Gb/s?

My answers,

At 01:04 AM 6/21/99 -0500, Thirion, Walt wrote:
What is ~10 Gb/s?

Ok, I would like to get the discussion really going. So, I'm going to propose a couple of issues. Hopefully, we'll get extensive reflector discussion, then I'll schedule a conference call for Wednesday to try to hash out differences.

A lot of the discussion revolves around the objective of whether we want to support a direct attachment to the WAN infrastructure by having a specification that maps 10Gig directly to an OC-192 like interface.

For the sake of discussion, my first objective is to nail down exactly which speeds we're talking about. There were a couple of numbers thrown out at the Interim in Idahao. The following table was taken from the web site at <>

Comparing Capacity
OC level Number of DS-3
circuits or STS-1 DS capacity ATM capacity POS capacity

tributaries (Mbit/s) (Mbit/s) (Mbit/s)

OC1 1 45 47 52
3 135 129 156
12 540 563 622
48 2,160 2,254 2,488
192 8,640 9,014 9,953
768 17,280 36,057 39,813

DS = Digital signal
POS = Packet over Sonet
STS =Synchronous transport signal


1. So, when we say ~10Gb/s, is there an actual number that we want and, if so, what is that number?
Yes, it is an actual number. The number is a data rate of 9.584640 Gbits. This rate is expressed after coding overhead (i.e. for 8b/10b coding the baud rate would be 11.98080 Gbaud). The reason this number is not identical to SONET transmission clock is because all SONET overhead has been subtracted. The SONET transmission clock is 9.95328 Gbits of which a few percent are coding overhead leaving a 9.584640 data payload. The numbers quoted for POS above are incorrect since they include the SONET overhead as data therefore not stating the actual line capacity.

2. Is there more than one WAN number that we should be considering? If so, what are they and what is the basis (standard) for the number?
The 9.584640 is the only number which is required. Adaption to OC-48 and OC-768 would be performed by inverse multiplexing and multiplexing respectively. The question is whether we need any special considerations to allow for multiplexing. This will depend on the PHY design not on the 10GMII interface rate.

3. Is the number a "raw" bit/baud rate or is it the net after subtracting some protocol overhead?
The 9.584640 rate includes any Preamble, SA, DA, Type, FCS, and IFG. It therefore includes protocol overhead, but not encoding overhead.

4. When we talk about a direct connection to the WAN, is there also an assumption that we will pick up the integrated management, i.e. self-healing Sonet rings?
The 10 GigE is not required to support self-healing or any other SONET management attributes. At the demarcation point between 10 GigE and a SONET ring a transcoding device will isolate the SONET characteristics from 10 GigE characteristics. The transcoder will allow SONET management on the SONET ring and Ethernet management on the 10 GigE ring.

An independent issue to the speed is whether 10 GigE should have SONET type management capabilities for self-healing and failure detection. These may be desirable objectives, but have little to do with the speed issue. The management issue is more important for dark fiber and dark wavelength applications than SONET mapping applications.

Walter Thirion

Level One Communications

Paul A. Bottorff, Director Switching Architecture
Bay Architecture Laboratory
Nortel Networks, Inc.
4401 Great America Parkway
Santa Clara, CA 95052-8185
Tel: 408 495 3365 Fax: 408 495 1299 ESN: 265 3365
email: pbottorf@xxxxxxxxxxxxxxxxxx