Cypress Preps Ethernet-Over-Sonet Chip

When Cypress Semiconductor Corp. (NYSE: CY) acquired Arcus Technology back in June 1999, it billed the event as "an aggressive push into the telecom/datacom markets." Now, nearly three years later, the company is about to announce the first product to come out of that acquisition.

It may have taken some time in development, but it looks as though Cypress is going to hit all the right buttons with its new chip. Codenamed POSIC2GVC, it will be able to map any protocol onto Sonet, using virtual concatenation (VC) and generic framing procedure (GFP) standards. (For a deeper explanation of VC, see PMC Pushes Sonet Silicon; for more on GFP, check out page 5 of Metro Multiservices Evolution).

These standards allow data traffic to be carried efficiently over the existing core network infrastructure, which is predominantly Sonet. That's a big deal, because carriers are desperately trying to save cash by preserving their existing investments in infrastructure.

Many of the big communications chip makers, including Agere Systems (NYSE: AGR), Agilent Technologies Inc. (NYSE: A), and PMC-Sierra Inc. (Nasdaq: PMCS), are making a big push into any-protocol-over-Sonet products, even though they are trimming product development in other areas (see, for example, Agilent Boosts Ethernet-Over-Sonet).

Indeed, at the Optical Fiber Communication Conference and Exhibit (OFC) last week, Agere announced that it had started volume production of its virtual concatenation framer, called TADM (see Agere Produces Sonet Framer Chip). An Agere spokesperson at the show told Light Reading that the TADM sits on line cards inside Cisco Systems Inc.'s (Nasdaq: CSCO) phenomenally successful ONS 15454 add/drop multiplexer, the box Cisco acquired from Cerent.

If true, Cisco's choice of Agere's chip is an important validation for the new Ethernet-over-Sonet framers.

Cypress appears to be playing catchup with Agere. Its chip started sampling to selected customers two weeks ago, according to Geoff Charubin, director of marketing for the company's datacom products division.

Superficially, the function of Cypress's chip seems similar to that of Agere's TADM: Each packs up to four gigabit Ethernet streams into a single OC48 (2.5 Gbit/s) pipe, using virtual concatention and what's called "oversubscription." There are likely to be substantial differences in the other features of the two chips, but, unfortunately, full details of Cypress's chip aren't available for comparison.

It's worth looking a little more closely at oversubscription -- what makes it possible to squeeze 4 Gbit/s worth of data into a 2.5-Gbit/s pipe. The theory behind it is that not every customer will fully utilize the capacity of its gigabit Ethernet channel. Data tends to come in bursts, so by holding back a burst of data momentarily, it can fit into a lull in the data on another channel. "It's kind of like what airlines do when they overbook planes," says Steve Perna, VP and general manager of PMC-Sierra's optical networking division.

Of course, when planes are overbooked, passengers with tickets sometimes get left behind. In the data world that means dropped packets. For that reason, not everyone is a proponent. "We're trying to provide carrier-class reliability," says Perna. "[Oversubscription] sounds like a good idea, but the implementation is very difficult."

PMC-Sierra's chip, the Arrow 2xGE, doesn't use oversubscription -- it packs the regulation two gigabit-Ethernet channels into an OC48, and uses the remaining space to carry circuit switched traffic.

— Pauline Rigby, Senior Editor, Light Reading
Page 1 / 2   >   >>
sg38 12/4/2012 | 10:42:37 PM
re: Cypress Preps Ethernet-Over-Sonet Chip Does Sycamore use this chip to pack 2 GigE into
one OC-48? I thought a while ago, SCMR made an announcement, right?

Would someone provide even more detail description on how Cypress chip works?

bad boy 12/4/2012 | 10:42:32 PM
re: Cypress Preps Ethernet-Over-Sonet Chip The way Virtual Concat works is something similar to IMA. So, if Cypress claims Virtual Concat, it works the same way:

A large payload like Gigabit Ethernet is split over channelized SONET say, multiple STS-1s or STS-3s. So a Gigabit Ethernet can take 21 STS-1s or 7 STS-3s. Then they are indexed as which is first and second and so on and sent over the network. On the receive side, you re-assemble them in the right sequence and extract your Gigabit Etherent back. The beauty of this approach is that it can go over legacy SONET, different SPEs can travel different path - like IP packet, and you can change bandwidth allocation.

I believe Cypress, PMC and Agere, all of them work the same way since this is standard based approach

opticaldude 12/4/2012 | 10:42:32 PM
re: Cypress Preps Ethernet-Over-Sonet Chip Lucent already has this featur in the DMX product line. They use 21 VC STS-1s for each line rate GigE. They can use any incriment from 1 to 21 STS-1s. If you use two full line rate GigE's on a OC-48 you have 6 STS-1's left for TDM or 10/100 services. Sycamore is a DWDM manufacture and does not have Sonet product. They are using a dual port per wavelength pack to get two GigE's on a lambda. It is still only WDM and not E-net over Sonet. Hope this helps.
bad boy 12/4/2012 | 10:42:31 PM
re: Cypress Preps Ethernet-Over-Sonet Chip "An Agere spokesperson at the show told Light Reading that the TADM sits on line cards inside Cisco Systems Inc.'s (Nasdaq: CSCO - message board) phenomenally successful ONS 15454 add/drop multiplexer, the box Cisco acquired from Cerent."


15454 is obviously not using Virtual Concatenation.

If you look at Cisco's site, it describes Gigabit Ethernet solution. The bandwidth allocation it describes matches what Agere's old generation chip TDAT offers - a lot of non-standard circuits like STS-6c, STS-9c and STS-24c.

If Agere's TADM chip is in 15454, why do they talk about non-standard stuff and not talk about Virtual Concatenation?

I saw Agere chip at OFC with huge fan on it. Also hear that it has problems. Does any one have insight?

opticaldude 12/4/2012 | 10:42:29 PM
re: Cypress Preps Ethernet-Over-Sonet Chip Lucent already has this in the DMX. Go to this page to see it.

sg38 12/4/2012 | 10:42:18 PM
re: Cypress Preps Ethernet-Over-Sonet Chip Thanks BB and opticaldude for your response!

Any pointer on where can we find out which standard document talk about Ethernet-Over-SONET?


bad boy 12/4/2012 | 10:41:14 PM
re: Cypress Preps Ethernet-Over-Sonet Chip There are multiple ways and standard for Ethernet-Over-SONET. For one, Ethernet frame can be wrapped into HDLC-PPP in a typical POS way. There are IETF RFC for that. There is also a new standard specific to this in ANSI called X.86 OR known as LAPS. Further, GFP (Generic Framing Procedure) is popular for Ethernet over SONET wrapping and it is ratified by ANSI and ITU.

Between these three, most, if not all, Ethernet over SONET traffic is covered.

opticaldude 12/4/2012 | 10:41:12 PM
re: Cypress Preps Ethernet-Over-Sonet Chip Don't forget about the forums that are out there like RPR and Metro Ethernet Forum. I believe that the Metro Ethernet Forum will win out in the end. It has more manufactures in it and is more standards based than the RPR. Look for RPR to be DOA.
bad boy 12/4/2012 | 10:41:11 PM
re: Cypress Preps Ethernet-Over-Sonet Chip OpticalDude,

I agree. I was more focussed on Ethernet over long distance and not just metro, and hence SONET.

What is better about MEF compared to RPR? What are the key technological enabling factors claiming MEF to win?

opticaldude 12/4/2012 | 10:40:57 PM
re: Cypress Preps Ethernet-Over-Sonet Chip BB,

I will not claim to be an expert on either, but from my point of view the biggest dif's are:

1. RPR only focuses on Ethernet over one ring(limited metro only). It does not look at Ethernet from Boston to LA. MEF looks at enet everywhere.
2. RPR does not build on existing standards, it is a new standard from the ground up. MEF looks at existing standards like Fast Spanning Tree, GRE Tunneling and VLAN's and others and incorporates them into its forum.
3. RPR is Cisco's push of their 15454/327 products. MEF has support from multiple manf and SP like SBC.
4. I donGÇÖt think SP will just throw out their existing Sonet networks and replace it with RPR. MEF will work with existing Sonet networks.
5. MEF looks at Virtual Concatenation of STSGÇÖs while Cisco (oops RPR) uses STS-1,3c,6c,12c and 24c. These can not be transported over standard Sonet networks.

Interested in others thoughts on this as well.

Page 1 / 2   >   >>
Sign In