Optical/IP Networks

Chip Startup Plans OC768

A California startup claims to hold the key to making switches and routers capable of supporting Sonet OC768 (40 Gbit/s).

Azanda Network Devices says it holds the key to making switches and routers capable of supporting Sonet OC768. The startup hopes to offer samples of a "traffic management chip" later this year that will support OC192 (10 Gbit/s) and OC768 Sonet rates. The chip, Azanda says, is a network processor designed to sit next to a switch fabric on a line card inside a switch or router. It will control the flow of data -- shaping, scheduling, and assigning priorities to specific types of traffic.

Azanda is among the first vendors to attempt to make a network processor chip that supports OC768. Network processors are off-the-shelf chipsets designed to perform specific functions -- such as packet filtering -- in switches, routers, and other network devices. For makers of networking gear, they provide an alternative to buying costly, customized ASICs (application-specific integrated circuits). It's a growing market inhabited by a troop of other newcomers (see Network Processors Proliferate).

Azanda says its claim to fame will be its ability to create a sophisticated chip out of mundane materials. It is eschewing fancy substrates and instead will focus on making a CMOS (complementary metal-oxide semiconductor) chip out of silicon. "Our focus is not going to be on exotic technologies in the early stages," says CEO Bidyut Parruck, whose last post was at Paxonet Communications Inc. (see Chip "Could Boost Metro Market").

Parruck says Azanda's found a way to supercharge its chips via new algorithms, some of which are dedicated to using memory more effectively. This, he says, is one of the major challenges of going to 40 Gbit/s. "Over 5 Gbit/s, bandwidth management is no longer easy on DRAM. You have to do something different with chips that manage bandwidth." So far, Parruck says, the company has seven patents pending on its techniques.

Parruck feels Azanda's focus will help distinguish it from other makers of network processors, such as ZettaCom Inc. (see Zettacom: Hurry Up and Wait). "They've taken so much on, it's going to be hard for them to deliver it all," says Parruck.

But Azanda faces competition from other sources, too. To succeed, it will have to contend not only with network processor startups but also with established players designing their own OC768 chips, including Lucent Technologies Inc. (NYSE: LU) and Nortel Networks Corp. (NYSE/Toronto: NT). Nortel's been demonstrating its OC768 chips for months now, although no announcements have yet been made of its availability.

Azanda will be up against flashy newcomers as well -- including CyOptics Inc. (see Vendors Prepare for 40 Gigabit Future), which has started sampling its OC768 chip, and Mintera Corp. (see Sycamore’s Stealth 40-Gig Strategy ). Challenges also will come from traditional chipmakers, such as Applied Micro Circuits Corp. (AMCC) (Nasdaq: AMCC), PMC-Sierra Inc. (Nasdaq: PMCS), and Vitesse Semiconductor Corp. (Nasdaq: VTSS).

Whether Azanda can succeed remains to be seen. In the meantime, it's all systems go for the startup, which opened its doors officially in May 2000. In addition to Parruck, the management team includes:

Azanda has approximately 45 employees, nearly all of whom are engineers. The startup has $10 million in series A funding from Bessemer Venture Partners, Commonwealth Venture Partners, Goldman Sachs & Co. (NYSE: GS), and Highland Capital Partners. Another round is planned for the March timeframe, Parruck says.

-- Mary Jander, senior editor, Light Reading http://www.lightreading.com

Mary Jander 12/4/2012 | 8:58:35 PM
re: Chip Startup Plans OC768 Any thoughts on just how important it will be for chip and component vendors (as well as equipment makers) to support OC768 next year? Some may lag behind. Will this necessarily be a detriment?
photonic314 12/4/2012 | 8:58:34 PM
re: Chip Startup Plans OC768 Please explain to me why there is such a push to 40 gb/s and 80 gb/s? My assumption here is that eventually they will like to apply OC768 to a WDM system, providing 40 gigs per channel...right?

Is ther something I am missing regarding the laws of physics and dispersion...?

How will 40 gig channels be put in a DWDM format and how many channels will there be....?

And won't it be expensive to then provision the 40 gig to multiple 2.5 gig customers at the access end?

Please help me understand why the need for speed when ultimately if I am not mistaken it all relates back to fiber capacity?
Petabit 12/4/2012 | 8:58:33 PM
re: Chip Startup Plans OC768 Hi photonic,

Since you asked a seriuos question, a serious answer.

Historically there has been a strong economic argument for increasing the bit rate. A OC 192 (10G) channel cost about 2.5 times a OC 48 (2.5G) channel. Therefore it was sensible to get four times the bandwidth for 2.5 times the cost. You also made huge savings in floor space, power consumption, management and operations.

Since the spacings of the channels was the same, you got way more bang for your buck.

OC 768 (40G) will cost a little more than 2.5 times OC 192, but not a much as four times. Since the spectral content of the signal is so much more, you have to space the 40G wavelengths further apart. So you won't get a four times increase in capacity on the fibre - but you will get some.

It all comes down to the cost of a managed bit, and increasing the bit rate has conventionally bee the route to reducing that cost. It will probably hold true for 40G, but 80G and beyond are less certain.

expressm 12/4/2012 | 8:58:29 PM
re: Chip Startup Plans OC768
Address: Express Marketing Concepts
3454 Saddleback Court
Port Orange FL 32119 USA


mailto:[email protected]

I understand there is no charge for info.

Thanks - Ken Bakker
pablo 12/4/2012 | 8:58:28 PM
re: Chip Startup Plans OC768
I don't think anyone doubts OC-768 is coming. The question is only when, what the initial business size is going to be like, and when it will ramp up.

It is also questionable whether traffic management is going to have any relevance at OC-768. The whole point of OC-768 is about keeping average link utilization down so one does NOT have to do advanced traffic management and STILL provide acceptable quality of service. So Azanda's traffci management claims at OC-768 are strange. OC-768 switch fabric? Important. OC-768 packet classification? Absolutely. OC-768 advanced traffic management? Not until 2004 or something like that.

I suspect Azanda, like so many others, are using OC-768 simply to drum up hype. Sure, it's on their roadmap somewhere, but right now and for the next few years everybody's bread is going to be buttered by the multi-billion OC-192 market.
dgindc 12/4/2012 | 8:58:19 PM
re: Chip Startup Plans OC768 DWDM (Analog Multiplexing) vs. OC-768 (Digital Multiplexing)

Remember when the modem market went through a similar issue in the early 90's ?

Modems went from 100 Baud to 300 to 1200, and finally to 2400. After that, the ANALOG technique of shifting phase, frequency, and amplitude to send multiple bits per baud enabled much faster speeds. No one realy questioned whether more signals could be sent per second just as most people believe that OC-768 and OC-3052 are technically feasible. The question is how will they be economically feasible when the analog method of dividing OC-48 and OC-192 waves can provide more capacity with less cost ?
dgindc 12/4/2012 | 8:58:18 PM
re: Chip Startup Plans OC768 Sorry for the error there, I meant OC-3072.
FrozenCanuck 12/4/2012 | 8:57:38 PM
re: Chip Startup Plans OC768 I would tend to agree with what Petabit wrote. If the cost of managing the bit is lower, there is an advantage. Besides just the technology, there are floorspace and thermal issues. 40G takes up less of both per Gb/s.

When you go from 10G to 40G, you have 16x more of a chromatic dispersion problem, and from 10G to 80G gives you 64x more of a problem. What about PMD? Also a big problem.

As you go to 40G, you essentially have a 4x larger spectral width, so as P. said, you have to be concerned with channel spacing being TOO close. At 80Gb/s this would get crazy. At some point these problems will drive more DWDM and less TDM.

Cool discussion :)
sjorrey 12/4/2012 | 8:57:35 PM
re: Chip Startup Plans OC768 The fiber problems you mentioned ,PMD,CD, spacing, etc... all can be over come by exactly the same methods mentioned that carried the modems to higher and higher speeds, and that is called "Optical FDM" As bit rates increase then higher order modulation techniques must be implemented to over come fiber non-linearities... and allow more bits per wave. Cost come down space comes down and is a very economical approach. www.kestrelsolutions.com
Cypher 12/4/2012 | 8:57:33 PM
re: Chip Startup Plans OC768 I am not here to advocate OC-768, but one benefit of higher TDM rates vs. higher DWDM rates is that the higher TDM solution (OC-768) reduces the number of Network Elements needed & theoretically simplifies network management.

However, I think the big push for OC-768 is history. People remember how Nortel created & subsequently dominated the OC-192 market & think/hope that it can happen again with the OC-768 (unlikely).
Sign In