x
<<   <   Page 3 / 4   >   >>
Abby 12/5/2012 | 1:27:02 AM
re: Why Sonet Chips Are Sexy Alok,

Great post! In fact, when you think of how much has been invested in creating multiservice boxes, it would be nice and profitable for whoever does it, to see a research report on this topic.
Peter Heywood 12/5/2012 | 1:27:02 AM
re: Why Sonet Chips Are Sexy Optiplayer - thanks for getting things back on topic.

Re your comment: "I'm hard pressed to think of a succesful start-up that used off the shelf silicon."

I know of at least two major vendors that have gone further than buying off the shelf silicon for Sonet/SDH equipment. They've outsourced the whole design, chips plus software (incorporating some of their own intellectual property). In one case, the resultant box has sold like hot cakes. In the other case, it's too early to tell.

These are major vendors, in the top 10 in terms of market share.

I hope to be able to reveal their names sooner or later.
particle_man 12/5/2012 | 1:27:01 AM
re: Why Sonet Chips Are Sexy When you converge things, the organizations responsible should also converge. This is difficult, political and slow. A couple of examples come to mind that a number of us are struggling with -

In the enterprise, is VoIP run by the voice or the data department?
In the carrier networks does an integrated sonet/DWDM solution get evaluated by the DWDM team or the SONET team.
Which organization evaluates an MSPP?

Technology is probably the easiest part of convergence!
OSXman 12/5/2012 | 1:27:00 AM
re: Why Sonet Chips Are Sexy Do you think Polaris has a chance in this environment? Sometimes if you stay still long enough, the market comes to you. Or not.
Kerry Davis 12/5/2012 | 1:27:00 AM
re: Why Sonet Chips Are Sexy >>We did just this at Onex Communications, made a chipset that melded together an Sonet ADM/Switch and a IP/ATM TM/SWITCH into a two chip solution<<

alok,

Believe me, I am sympathetic to everything you are saying here and I was right there at Polaris when Onex was going through that development effort. However, the approach used by Onex and Polaris is actually not what I am suggesting at all.

That approach will result in an office with multiple multiservice boxes with their individual multiservice switching fabrics interconnected by intraoffice trunks. Besides not scaling to the customers needs and the management difficulties, you have all the development headaches you very accurately mentioned.

The alternative is to create a single multiservice switching fabric (either integrated right into the I/O card or a standalone product in itself or both) that scales one to one with, and is shared by, multiple single service I/O boxes. This rules out a chip that scales using a centralized multistage architecture like the one Onex created. The result is an office with many single service boxes sharing a common scalable switching fabric (as opposed to many interconnected multiservice boxes )which functionally accomplishes the same thing. The carrier can add I/O cards and fabric as needed. But you have to quit thinking about a redundant Clos architecture to make this work and this is very hard for many to do.

K
Drew Lanza 12/5/2012 | 1:26:59 AM
re: Why Sonet Chips Are Sexy Tony Li is right that up until, well, right about now, the major vendors like Nortel, Fujitsu, and Lucent have had a lock on transmission gear and have been able to sustain very large margins (i.e. prices) on those products.

My friend, Farooq Raza, the founder of Mayan Networks, did a good study of the erosion of SONET prices in the 90's. Not surprisingly, he discovered that they eroded very little - maybe 15% per year. Not what you would expect from electronics being sold in moderate volume and supported by the dynamics of its underpinning silicon and optoelectronics.

I don't think we'd be talking about the commoditization of transmission silicon (SONET, metro Ethernet, ADM's on a chip) if the incumbents still controlled the market.

The interesting change here is the emergence of Cisco as a force to be reckoned with in the transmission space (amped up after they acquired Cerent and, perhaps, by the recent HFR announcement) and, of course, Huawei in China.

I posted a couple of years ago that I thought Cisco seemed to have made a concrete decision to go after the major transmission players by deliberately eroding the price of transmission gear. Of course, Cisco has to maintain their high margins, so they have done this by aggressively supporting the development of integrated silicon and optoelectronics to disrupt the cost of building 10 Gig (and slower) linecards. This strategy will certainly enable Cisco to crush the majors, but it remains to be seen to what extent it will help Cisco to grow their revenue.

The Chinese are motivated for different reasons, but the outcome is the same. It seems silly to pay tens of thousands of dollars for 10 Gig linecards when they can be built up from XFP transceivers, monolithic serdes/framer/protocol chips and NP's, each of which only costs a few hundred dollars. Of course, China already has more DSL lines than any other country in the world. By 2006 they will have more homes and offices connected over broadband than any other country, and they are only beginning.

The other twist to integrated silicon is that the driver software becomes very sophisticated and must be supplied by the chip vendor. This has all sorts of ramifications on funding and time to market, but it also may enable the Asian (especially Taiwanese) ODM's to build reliable, feature-rich SONET gear that is smaller, faster, and disruptively cheaper than anything the majors have ever delivered or envisioned.

I'll be honest, I don't know where all of this leads. In the end, network demand comes from the edges of the network and that means ripping up roads to lay fiber to connect to buildings or to push HFC networks closer to homes or to saw up CSA's to distances more like 5kft, and making DSL go faster often means replacing copper CPE and CO equipment. These are always expensive and time consuming processes that dictate the slow pace of network evolution. Even if SONET and metro Ethernet were free, would it have an impact?

I believe we will see an explosion in low cost SONET and metro Ethernet gear as Peter suggests. Clearly, the technology is there to push it along. But wether the customer pulling it through will impact this business or our lives significantly is something I've never figured out how to gauge.

As for the difference between hard news and soft editorials, I've given up. Since I'm always on the go, I try to get my news from 4 sources every day. NPR in the car, the Yahoo news page in the office, and the O'Reilly Factor and the Chris Matthews Show at home at night. NPR leans left, Matthews and Yahoo seem to be in the middle, and O'Reilly comes from the right.

It has become increasingly interesting to note that at least once a week I find myself in a situation where two of these sources directly and diammetrically contradict each other on something that I would consider as 'hard' news. One says the world is round. The other says it is most definitely flat.

It has become impossible for most of us to separate opinion from fact, especially in the political arena. Caveat Emptor.

Drew Lanza
alok 12/5/2012 | 1:26:59 AM
re: Why Sonet Chips Are Sexy I hope polaris does well, they have a great h/w, s/w team that put in a lot of hours to get the product done! The issus is with the market, when the market shifted away from god boxes, they focussed on the TDM ADM/SWITCH market place with a product that offers higher density and easier manageability than the competetion, ie tellabs. Tellabs got back into the game by aquiring occular networks. I think the carriers are hesistant to buy from startups and use polaris to drive down prices from tellabs and other suppliers. Someone should aquire them and then make a go of it or they should parter with a big boy. On their own I dont see them succeeding.

My two cents!
Alok

paolo.franzoi 12/5/2012 | 1:26:56 AM
re: Why Sonet Chips Are Sexy
Drew,

I think what you are saying is:

1 - Look at what is happening in the SDH market, particularly in South Asia. When Huawei and ZTE show up, the products are sold for less than the BOM costs of the major suppliers.

2 - This change can come to SONET networking in the US, if the vendors can find a) channel and b) meet all the standards.

I guess my commentary is that this has little to do with the ASIC versus merchant discussion. Each product needs to understand what its competitive advantages are. Then, R&D must be focused to create those competitive advantages. If I am an outside looking in this either is a dramatic cost change (at least 30%) or a functionality that can not be matched.

The issue with merchant is: Can I do the job with a small enough staff that I can accept lower gross margins and still have a good business?

seven
alok 12/5/2012 | 1:26:52 AM
re: Why Sonet Chips Are Sexy Kerry,

>>>Believe me, I am sympathetic to everything you are saying here and I was right there at Polaris when Onex was going through that development effort. However, the approach used by Onex and Polaris is actually not what I am suggesting at all.

That approach will result in an office with multiple multiservice boxes with their individual multiservice switching fabrics interconnected by intraoffice trunks. Besides not scaling to the customers needs and the management difficulties, you have all the development headaches you very accurately mentioned.

The alternative is to create a single multiservice switching fabric (either integrated right into the I/O card or a standalone product in itself or both) that scales one to one with, and is shared by, multiple single service I/O boxes. This rules out a chip that scales using a centralized multistage architecture like the one Onex created. The result is an office with many single service boxes sharing a common scalable switching fabric (as opposed to many interconnected multiservice boxes )which functionally accomplishes the same thing. The carrier can add I/O cards and fabric as needed. But you have to quit thinking about a redundant Clos architecture to make this work and this is very hard for many to do. <<<<

Kerry,

I guess I do not subscribe to this version of how things should work. The main advantage of having "one" multiservice box ( A big one at that since the chip set scaled to over 1 Tb/s total B/W) is management. The OPEX costs would be much more than the CAPEX costs over the life of the products, this is more so if you have different boxes from different vendors, and we wanted to reduce this to the minimum. You would have one management interface/console to set up/teardown TDM connections in a hitless fashion, provision new VPI/VCI connections, update IP tables etc. Also having one fabric allowed you to tune the granularity of the traffic between TDM/IP and ATM and gracefully migrate from one to the other, with no hardware upgrades required.

What I see happening in the future is that SONET will be treated more as a BIG FAT PIPE - OC-192 and above without being concatanated from smaller tributaries and data being stuffed down that pipe,
ie LCAS is going in that direction, giving you the
50 ms restoration etc as a transport layer. The data layer will be ethernet frames and MPLS for layer 2 1/2 switching and offering other services like ATM/FR emulation etc. Other services will be offered on top of the MPLS layer. The commoditization of the SONET chip sets and subsequent drop in transmission costs will only hasten this model. SONET works well and the carriers are very comfortable with this. I think polaris also realizes this and is hence pushing
dense/easily managed SONET ADM/CrossConnects as their value proposition.

So my big question to the community is "what are the interesting problems to be tackled in the networking industry, for if the above model is what is going to happen, as I believe it is most issues are aldready solved"

thanks for listening
alok
Scott Raynovich 12/5/2012 | 1:26:51 AM
re: Why Sonet Chips Are Sexy Tony, your analogies are not accurate, you are talking about when editorial is advertising a THIRD PARTY's product.

Media companies promote themselves all_the_time ... and they do it through editorial product.

You don't think the New York Times doesn't promote it's own columnists? Does FOX News not tell you what's coming up next on FOX News? Are you saying Dan Rather doesn't tell you during CBS evening news what they're going to show on 60 Minutes? Are you saying that ESPN News doesn't tell you what they showed on ESPN Baseball the night before? Of course they do, because they own it.

That's called cross-promotion, it happens every day in the media world.

Like I said it's not like we're trying to sell you Ginsu knife or a Juniper router. We're pointing to our own research. And we're not going to stop doing it. Sorry.
<<   <   Page 3 / 4   >   >>
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE