Comms chips

PMC Pushes Sonet Silicon

PMC-Sierra Inc. (Nasdaq: PMCS) today announced a chip that sandwiches two gigabit-Ethernet channels into an OC48 Sonet wavelength (see PMC-Sierra Launches Arrow).

The chip, called the PM5397 ARROW-2xGE, implements recently finalized virtual concatenation (VC) and generic framing procedure (GFP) standards. Both standards have been devised to offer more bandwidth-efficient ways of packing Ethernet traffic into a Sonet/SDH transport network.

"Carriers are spending less, but what they are spending on is Sonet," says Steve Perna, VP and general manager of PMC-Sierra's optical networking division. This realization has encouraged PMC to place more emphasis on its Sonet developments while putting its packet processor work on the back burner (see PMC-Sierra Pulls Packet Silicon)

To understand what's so cool about virtual concatenation entails looking at how data is transported over existing networks. Simply put, the problem is that Sonet channels are the wrong size for carrying Ethernet traffic. In "concatenated" Sonet, the channel is simply treated as a fat pipe. Putting a single gigabit-Ethernet channel (1 Gbit/s), for example, into an OC48c (2.5 Gbit/s) wastes 58 percent of the bandwidth.

Table 1: Sonet virtual concatenation service applications
Service Bit rate Without VC With VC
Ethernet 10 Mbit/s STS-1 (20%) VT1.5-7v (89%)
Fast Ethernet 100 Mbit/s STS-3c (67%) STS-1-2v (100%)
Gigabit Ethernet 1 Gbit/s STS-48c (42%) STS-3c-7v (95%)
Low-speed ATM 25 Mbit/s STS-1 (50%) VT1.5-16v (98%)
Fibre Channel 200 Mbit/s STS-12c (33%) STS-1-4v (100%)
Fibre Channel 1 Gbit/s STS-48c (42%) STS-3c-7v (95%)

An improvement on this is so-called "channelized" Sonet, which allows carriers to carve up capacity in an OC48 link, say, into units of STS1 (51.4 Mbit/s), STS3 (155 Mbit/s), or STS12 (622 Mbit/s). But it's not possible to mix and match these units because there's no guarantee that the same combination of units will be available at both ends of the network, if at all.

Enter virtual concatenation. Its job, in a nutshell, is to "right size" Sonet channels, to allow more efficient provisioning of bandwidth.

PMC's new chip takes two gigabit-Ethernet channels and independently maps each of them into a STS48/STM16 channelized payload, using VC and GFP standards. Each gigabit-Ethernet channel most closely fits into an OC21v (also denoted STS3-7v , or seven units of STS3, in the table above). What's more, the capacity that's left over (two lots of STS3 worth) can be utilized to transport voice channels.

"This device allows you to physically have access to the gigabit Ethernet streams before they get mapped onto Sonet," Perna explains.

Transporting Ethernet over Sonet using VC is a better option than native Ethernet networks, Perna contends. For starters, it won't require carriers to get a forklift upgrade to their existing networks, which are predominantly built on channelized Sonet. To support VC, a carrier only has to upgrade the equipment on the ends of the connection. What's more, Sonet offers features that native Ethernet doesn't, such as bandwidth guarantees, and redundancy in the event of a cable break.

It's worth pointing out, however, that PMC is not the only chip maker that thinks support of Ethernet over Sonet is a smart move. In October, Agere Systems (NYSE: AGR) also unveiled what it dubbed "an add-drop multiplexer on a chip", which supports VC and GFP standards (see Agere Offers Single Chip ADM).

The chips from Agere and PMC are quite different, however. Agere's chip integrates a pointer processor with an STS1 granularity crossconnect. PMC's chip, on the other hand, doesn't include the STS1 crossconnect, but does have onboard SerDes allowing it to interface directly to an optical module on one side and a backplane on the other, if desired. "This is one of the highest integration devices we've ever produced," PMC's Perna claims.

It’s also worth noting that system developers like Sycamore Networks Inc. (Nasdaq: SCMR) have had equipment for more than a year that crams two gigabit-Ethernet channels into a 2.5-Gbit/s wavelength (see Gigabit Ethernet Goes Carrier Class). Sycamore, however, had to develop its own silicon to achieve this feat. The off-the-shelf chips now being offered by Agere and PMC make it easier for other system vendors to catch up.

— Pauline Rigby, Senior Editor, Light Reading
Page 1 / 2   >   >>
opticaldude 12/4/2012 | 10:42:31 PM
re: PMC Pushes Sonet Silicon Lucent already has a product on the market that does this. The DMX and DMXpress uses VC of STS-1's to transport from 1 to 21 STS-1's of e-net. This product has been on the market for nine months. You can also take the STS-1VC-X across any TDM fabric that uses STS-1's. This means that anywhere you can get a STS-1 in the world you can get a LAN 10/100/1000. No more DS1/DS3 to transport your LAN. Also, RPR only looks at transporting a packet in a single ring. This VC style gives you a LAN connection anywhere you can get a STS-1. I think that this kills the RPR standard(BIG TIME). talk to me...
hogdog 12/4/2012 | 7:27:43 PM
re: PMC Pushes Sonet Silicon Who cares, we all know that PMC Sierra canned everyone who designed this supposed Brilliant Chip

Where did all of the Abrizio folks go?

Oh yah, they were fired!!!

Look out for Cisco!
(o)(o) 12/4/2012 | 7:27:43 PM
re: PMC Pushes Sonet Silicon "a chip that sandwiches two gigabit Ethernet channels into an OC48 (4.8 Gbit/s) Sonet wavelength"

I think that 4.8 Gbit/s should read 2.4 Gb/sec, no?
And what exactly does this chip have to do with Sonet "wavelengths"?? We're talking strictly electrical here no?
hogdog 12/4/2012 | 7:27:42 PM

8point 12/4/2012 | 7:27:42 PM
re: PMC Pushes Sonet Silicon i think that that technology is outdated by now. dsl over 56k modem is the way to go. either that or carrier pigeons. is that what cisco is doing with the abrizio team?
hogdog 12/4/2012 | 7:27:41 PM
re: PMC Pushes Sonet Silicon (O)(O)

Wall mart is having a sale on Mattel 2 way radios. They are walkie talkies that can be used when playing hide and go seek.

So you don't get lost!
gea 12/4/2012 | 7:27:34 PM
re: PMC Pushes Sonet Silicon This was a good article. Paulina Rigby seems to understand the implications of virtual concatenation.
Of course, there has been silicon available for a while that could pack 2 GbEs into an OC-48 by using various kinds of tricks, but you can't hand off those GbEs to someone else's gear without digging out the GbEs.
Virtual concatenation will change the shape of SONET networks and, possibly, make running native GbE much less desirable.
poster 12/4/2012 | 7:27:32 PM
re: PMC Pushes Sonet Silicon agreed. good job. see, I can give credit where credit is due. and since I agree with gea's point that this standard could prove to have profound effects on how ethernet is transported in the network (over SONET as opposed to natively, or via one of the numerous proprietary methods being developed), it would have been nice to see a big picture perspective of how this might effect different players in the industry. there are a large number of companies who are working on proprietary methods to do this same thing, and we've seen how much engineering and marketing effort they've all put into their respective approaches.

let's explore what this means to the MEF, RPR, and the proprietary ethernet-over-X efforts. what about the 10GbE camp, and the existing SONET players...? who else is making chips that are X.86-compliant? there are implications for all. this deserves a report!

BTW: this was designed into ITU G.709 digital wrapper as well.

gea 12/4/2012 | 7:27:30 PM
re: PMC Pushes Sonet Silicon Poster? Are you feelin' allright? Is that you saying something's GOOD?

As for me, a major stumbling block is the need for some sort of nice way to move GbE at 2.5 gig. One GbE/wavelength isn't quite enough to justify eating up a wavelength.

While I think this may have an impact on the Metro Ethernet Forum (who would likely want to push out GbE native), I don't think the same applies to RPR. Even with STS-3c-7v (or, more likely in off-the-shelf silicon, STS-1-24v), UPSR and certainly BLSR-style TDM protection doesn't really fit too well (ie, with VC you still only get two GbE circuits at OC-48). Thus we'll still need either MPLS or RPR to do some nice packet-level protection.

As for 10GbE, this may actually be a good thing. There's a big gap between 1 gig and 10gig, and this virtual concatenation approach may be what allows GbE metro carriers (after they start making money, of course), to move nicely between the two. It certainly allows them to use OC-48 transponders (and SONET framing...a big plus) to move GbE traffic.

I still think, however, that we may see a 2.5GbE standard in the near future, modeled after the 10GbE MAN/WAN standard (that uses SONET frames).
femtokid 12/4/2012 | 7:27:28 PM
re: PMC Pushes Sonet Silicon the 48 in OC48 does not mean 4.8Gb.
its an aggregate of OC1 which is 52Mb

times that by 48 and wheeehay u get 2.5G u donkey.
this is a good chip with good implications for the industry. to late to save me tho.

why do i bother

Page 1 / 2   >   >>
Sign In