RPR Moves Forward
Today roughly 80 percent of the group, consisting of about 100 individuals from system vendors, chip makers, service providers, and academia, accepted the joint proposal from Cisco Systems Inc. (Nasdaq: CSCO) and Nortel Networks Corp. (NYSE/Toronto: NT) known as “Darwin”.
"After long debates on the principles of the standard, we're finally at a stage where we can continue to the next and final phase -- writing down the actual standard," says Gady Rosenfeld, director of strategic marketing for Corrigent Systems.
This meeting is particularly important because for months observers have questioned whether or not the group would be able to achieve its goal of having a single draft by the end of the January meeting. Back in November it looked as though the two main groups -- “Gandalf” led by Cisco and “Alladin” led by Nortel -- were at a standstill (see RPR: Deadlock Ahead?). But now the two have come together and submitted a joint proposal: Darwin (see The Darwin Approach: Evolution of Resilient Packet Ring).
While Darwin has been marketed as a joint effort between the Gandalf and Alladin supporters, Fredric Thepot, an independent consultant attending the meeting, says that the new proposal is simply a redrafted version of Cisco’s Gandalf proposal.
“If you take the two texts, what you see is the Gandalf proposal with a few things thrown in from Nortel,” he says.
First, consider bandwidth management. Darwin, like Gandalf, doesn’t mention virtual output queuing, which controls congestion by limiting the number of packets entering the ring. But it was mentioned in the Alladin proposal. John Hawkins, senior manager of marketing for optical Ethernet at Nortel downplays this omission and explains that since queuing is not part of the MAC (media access control) layer, which the group is trying to define, it was left out of the standard.
“You can still do virtual output queuing using Darwin,” he says. “Even in the Alladin proposal it was an option.”
Hawkins also points out that in terms of bandwidth management, both Gandalf's congestion management and Alladin's congestion avoidance schemes are included in the proposal.
“Each group was espousing the virtues of their own approach," he says. "But there are networks that can use either method depending on the service provider's requirements. We decided we ought not preclude one or the other.”
“It’s like the difference between driving a car with an automatic transmission or one with a manual transmission,” adds Jeff Baher, director of marketing, Metro IP Access Business Unit for Cisco.
The Alladin group also seems to have made big concessions when it comes to packet formatting. Hawkins agrees that the Darwin packet format has the same look and feel as Cisco’s originally proposed spatial reuse protocol (SRP), but he says that additional fields have been added, like a header error checking field, which is required for more carrier-class applications. He also admits that more work is needed in this area.
“There are some complications in bridging between rings," he says. "And there will likely be other complications that come up that we haven’t even thought of yet. But this is a starting point.”
One area where Cisco seems to have compromised is in the protection scheme. The Gandalf proposal recommends "packet wrapping" as the default, with "packet steering" as an option. Packet wrapping is a scheme that calls for packets to simply turn around and go the other way around the ring when they hit a disconnection. Steering calls for all nodes to be notified of a failure and would direct traffic away from the fiber break. Alladin supporters proposed that the standard include steering only. In the end, the Darwin proposal uses packet steering as the default with wrapping as the option.
“The most important thing is to meet the 50 millisecond restoration time requirement,” says Cisco’s Baher.
In the end, members of the working group are satisfied that there will be a single draft to work on from here on out. But Thepot suggests that it has come at a particularly high cost to Nortel's installed base of customers. He believes that Cisco, which also has a large installed base, will have an easier time adapting its technology to the Darwin proposal than Nortel.
“To me it looks like Nortel lost in all of this,” he says. “Yes, it’s true that Cisco isn’t exactly compatible to Darwin, but it will be an easier migration. Nortel will have to go back and implement a lot more things, and I don’t see it being easy.”
Now that a preliminary draft has been accepted, editors will work on drafting a formal document. The edited draft will then be ratified at the next meeting in March and voted on again in July. A final version of the standard won't be ratified until early 2003, says Hawkins.
— Marguerite Reardon, Senior Editor, Light Reading