Interview with NMS Communications' Brough Turner
Editorial director Joe Pavlat spoke with Brough Turner, SVP, CTO, and cofounder of NMS Communications not long after Brough returned from meeting with carriers and telecom equipment manufacturers in Japan, China, and South Korea.
Joe: When you start to look at the numbers, for example, at the increased bandwidth that IPTV is going to require, two things jump out: One, the carriers are going to have to spend a ton of money developing infrastructure and, two, while the carriers used to control the transport and much of the content, more and more we see content providers (Google, YouTube) getting increasing chunks of revenue. There is less revenue available for the carriers, yet they are going to have to upgrade their networks to compete.
Brough: Verizon is seeing its fixed line voice revenues decline as people switch to mobile. If customers stay fixed, they are flipping to cable TV phones so the only thing Verizon can do is go after the cable TV‚Äôs turf. Verizon can certainly deliver IPTV over fiber because they have plenty of bandwidth there. They just take a piece of the available bandwidth and dedicate it for TV. AT&T has a bigger problem because they are doing fiber to the neighborhood and very high-speed DSL to the home, so they can only support two TV channels at a time. That means they can support two TV sets in the home right now, and they are trying to get up to three.
Joe: You tell COMcast Cable‚Äôs server what program you want in HD, for example, and they pipe that content over the cable.
Brough: Right. You get one per home. And you get fiber to the neighborhood or fiber to somewhere nearby. There is plenty of capacity on the fiber. You send a separate dedicated stream to each television set, and if it is AT&T you have to do the last few hundred feet to the last kilometer over very high-speed DSL, with the result they are limited to two TV sets per house.
Joe: But that suggests smart servers select content and are going to be in every neighborhood?
Brough: A ton of equipment companies are going after IPTV because the [Regional Bell Operating Companies] RBOCs purchases are so large, and behind that there are server farms and so forth. I‚Äôve had extensive discussions over pricing and affordability and paying for it all.
The reality is if you took an Internet-centric approach then you would do peer-to-peer streaming distribution such as Joost or any of the emerging peer-to-peer TV services, at which point you would automatically get caching in the apartment building or the neighborhood or wherever. Seldom-accessed content would come from a server somewhere across the Internet, and frequently accessed content would come from somebody close by, without using any of the backhaul. There are a lot of approaches, but until there is enough Internet bandwidth that it just happens over the Internet, there is a window for the RBOCs to try and stay alive while being assaulted by Mobile and by Cable.
Joe: But IPTV is going to place much strain on networks already running near capacity, and in some cases they are experiencing problems already.
Brough: Typically the IPTV is not running over the Internet path. I thought you were talking about Internet bottlenecks‚Ä¶
Joe: I was thinking of Metro Area Networks in general.
Brough: Once you get out of the neighborhood there is no capacity problem on anybody‚Äôs backbone; there is no congestion; there are no QoS issues. The backbone is an unregulated open market free-for-all with as much capacity as you want and competitive pricing. If you can get from your home or your business to the backbone, for example, from our building in Framingham to 1 Summer Street in Boston, you can buy Gigabit Ethernet connectivity from Cogent for $1,000 a month for one gigabit fully loaded. That would be enough to carry Framingham, Newton, and three other cities. It‚Äôs only $1,000 a month and offers connectivity anywhere in the world‚Ä¶
Joe: If you can get to the backbone‚Ä¶
Brough: I know the folks at M.I.T. who spent nearly a million dollars two years ago to buy 100 fibers in a ring between Cambridge and Boston, that is, basically a ring from one end of the M.I.T. campus to 1 Summer Street in Boston out Commonwealth Avenue through Back Bay to Boston U. and then across the river to the other end of the M.I.T. campus with 100 fibers in a loop, only two or three of which they are using. They pay 10 dollars per megabit per second per month using Gigabit Ethernet from Cogent ($1,000 a month). Once you get fiber to where the backbone is, it is a good deal.
Joe: Certainly IPTV was a big thing at NXTcomm and also Huwai‚Äôs much anticipated AdvancedTCA products, including quad core Opteron ATCA CPU, which was quite something to behold.
Brough: While NXTcomm was happening, I was in Korea, China, and Japan. In Japan, Docomo for years has been pushing AdvancedTCA pretty hard. I asked about AdvancedTCA at KDDI and SOFTBANK. They are both interested and think it is a good idea, but in the end they only care about price.
In China, China Mobile has joined in pushing AdvancedTCA on Huawai.
I was talking to a company in China that does ‚Äúlawful‚Äù intercept. They are in a niche market and design systems with four E1s. Then they spread thousands of them all over the country. These are dirt-cheap things. This company told me they cannot see justifying full-size AdvancedTCA, but they are interested in MicroTCA, and they know they need to pay attention because China Mobile is telling them they should be using AdvancedTCA. So there is a lot of recognition of AdvancedTCA in all three countries.
There is general agreement that certainly for compute-intensive applications like softswitches or Home Subscriber Servers (HSS), price crossover has happened, and AdvancedTCA is not only more appropriate than 1Us and cheaper than CompactPCI, but interestingly, in two cases I was told that it is considered lower cost than blade servers from IBM, HP, or Sun.
Joe: I have not looked at Sun or HP, but I have looked at the pricing for BladeCenter T, and it does appear that AdvancedTCA has a 25-30 percent cost advantage, depending on the estimates you make and volume.
Brough: The principal argument for these three companies was not about the first purchase price but about the annual maintenance fees
Joe: Total cost of ownership?
Brough: Yes, apparently the annual maintenance fees from IBM, Sun, and HP are outrageous.
Joe: Sure. They don‚Äôt want to sell you just the hardware; they want to sell you the software, the building, the swivel chairs, and the people.
Brough: The Tier 1 NEPs and the major operators are only looking at full-size AdvancedTCA in Japan. I was told that Docomo does not like edge connectors so they will never use AdvancedMCs. However, the guy that said that also said: ‚ÄúWe sell in other parts of the world so we are very interested in what happens with MicroTCA.‚Äù
Joe: The great prototype question is: If somebody wants to buy 10,000 boards a month, are they going to want AdvancedMCs or are they going to want that logic residing on the base board?
Brough: I think it is very clear from people I have talked to that if they want a full-sized AdvancedTCA solution the most they would accept would be an AdvancedMC plugged into an Intel blade or something like that; they would not accept four AdvancedMCs plugged into a carrier except for software development.
Joe: That is certainly what I am hearing.
Brough: But I laid out a strategy that said, ‚ÄúThere is all this flexibility and scalability in AdvancedMCs. Any custom configuration you want me to produce for you in x number of months with an appropriate volume commitment is possible,‚Äù and that seemed be acceptable.
Joe: Clearly AdvancedTCA growth is in Europe and Asia and to some extent in South America. North America is pretty flat. The smaller companies that are only playing in North America are not enjoying success.
Brough: Other things that I saw were extremely low cost MicroTCA chassis. I spoke to people who are specifically interested in such chassis, both in Korea and in China.
Joe: MicroTCA has been criticized for being an expensive solution even before payload cards are added. How did they achieve low cost?
Brough: In the very low-cost MicroTCA chassis, the MCH functions are built into the backplane; it is an active backplane. There is an open frame power supply, the cheapest commercial supply they could buy, which is wedged in the back. Four mid-height or full-height AdvancedMC modules plug in from the front and the whole thing is a 1U chassis. There‚Äôs no High Availability (HA) support or story, but it‚Äôs a solution that is flexible.
Joe: There is a chicken and egg cost issue with the radial power supply lines and the complexity of the MCH. However, when the volumes get high enough, a lot of that can be integrated into silicon.
Brough: In the Media Resource Function (MRF), which in the IMS architecture is the universal media box, it seems that just about everybody has agreed there are going to be two versions. There is going to be a high-capacity, low-cost-per-port voice-only MRF, which is like an announcement server that is used by the softswitch and everybody else, and there is going to be a more expensive, more flexible, multimedia MRF.
The voice MRF is probably not an opportunity for NMS because the major equipment providers are each building it themselves in-house. In fact, anything very high volume they will build themselves. One major equipment provider showed me a chart (on a napkin during dinner) illustrating their different products in the AdvancedTCA ecosystem. The chart showed how they had started a couple of years ago and where they would be in 2008. In 2005 they were using chassis and CPUs from Intel and switches from ZNYX. By 2008 they will be building their own chassis, they are already building their own CPUs, they will still buy switches from ZNYX, but they will be building their own hard drives.
One thing I heard was: ‚ÄúIf we are going to spend more than ten million dollars a year on a particular component or a particular thing at a board level or chassis level then we have to think seriously about building it ourselves in China.‚Äù
So NMS has a big opportunity in multimedia MRF with the Tier 1s, and we have significant opportunities particularly in mobile video related stuff and probably over time in AdvancedMCs and MicroTCA for a wide variety of other people.
Joe: Do you see NMS in the future being primarily associated with voice?
Brough: No. We got involved with Docomo, NEC, Fujitsu, and NTT Commware with mobile video in 2001 and did nine different NRE contracts over three or four years and came up with a product line called Video Access and subsequently with video transcoders, video servers, and a variety of video development environments. If you just want voice, the reason you would buy from NMS is you know you can get the video later and if you want video, we are there with both the media and the signaling and a complete platform, as opposed to the pure-play video guys who have pieces of the solution but not the whole environment.
Joe: You have been a proponent of mobile voice, which has been a fundamental driving force in emerging economies. Will video have the same impact? Or will it simply be ‚Äúnice to have‚Äù?
Brough: Mobile voice is penetrating the developing world, and if you look at what people want next, they want different levels of Internet connectivity. Video may not be that important in a developing country. We will hit three billion in 2008 who physically own their own handset. I saw studies in Sri Lanka of Sri Lanka, India, Pakistan, Malaysia, and the Philippines looking at the lower fifth income level people in those countries. These studies showed 94 percent of the individuals in that income bracket had made a phone call within the past couple of months, and nearly 70 percent could get to a phone within five minutes. That is very different from the situation five years ago. In 2000 half the people in the world had never made a phone call. Now 94 percent of the people in the bottom-most economic rung have made a phone call in the last couple of months, and 70 percent of the people at the very bottom rung of society are within five minutes of a phone if they need to make a call.
Joe: You have spoken about the importance of primarily voice and secondarily Internet access as a way to get access to information. Do you see voice access to databases as part of the equation in the future?
Brough: Voice access is important for those who can‚Äôt get Internet access, or are not literate, or use a language not understood by the SMS interface or the keyboard. A lot of voice access to information is certainly happening. Speech recognition is still not Star Trek level, but it keeps getting better and better. Four or five months ago I flipped over on my Verizon cell phone to using GOOG 411, which is Google‚Äôs information service. It is completely speech recognition driven and extremely good, and it is also free ‚Äì much better than calling 4-1-1 on my Verizon phone. Speech recognition has its problems, but it keeps getting better and better as Moore‚Äôs Law drives the number of MIPS that you can apply to it.
Joe: Handheld devices are generally lousy for surfing the Net. How do you see that developing as we get better mobile access to the Internet?
Brough: In 10 or 15 years it will be easier, but the prospects for the next couple of years are not that great.
Joe: iPhone notwithstanding.
Brough: Right. If you prepare to look out a decade or so, the available bandwidth on 3G HSDPA and HSUPA ‚Äì the sort of networks we‚Äôll see in the next 24 to 36 months ‚Äì you get to the point where you have more than a megabit down and several hundred kbits per second up‚Ä¶.
Joe: Will the devices themselves become larger?
Brough: You have a couple of choices there. One, you can use new interface technologies like the accelerometer stuff that is used in the Wii interface, the sort of stuff the iPhone is experimenting with, including speech recognition. Two, if you want a keyboard, there are early examples of keyboards that are an LED projecting a keyboard on a table. You type, and it sees where your fingers are touching the table.
There is a lot of bizarre stuff if you are willing to talk 10 years out. I am relatively confident that something will happen, but the smallest device you get might be able to turn into a computer (in the sense you think of a laptop) by using an LED projector and some sort of projected keyboard. A lot of things could happen in 10 years, but they are not going to happen in two years.
The user interface is the big bottleneck. You can think about all the interesting user interface technology, but when someone is going to get it right is arguable.
Joe: NMS has been involved in mobile video for some time.
Brough: The biggest opportunity for NMS is video. We got involved back in 2001 with the early Docomo mobile video systems, and at that time the idea was people were going to do video telephony, they were going to call each other, and that is what the 3G-324M video telephony standard was designed for.
It is very interesting technology, but what we find people are actually doing, if it‚Äôs live person-to-person, it is what we call see what I see video, that is; it is a two-way audio connection, but a one-way video connection. Typically someone in an interesting place is using his or her phone as a mobile webcam, for example, look Mom I am climbing Mt. Fuji. So, your mobile phone becomes a mobile webcam. You are doing two-way audio to talk with someone, but it is one-way video. It does not matter what technology you use ‚Äì 3G-324M or you can use conventional voice plus video over IP, which is what the recently announced AT&T video sharing is.
NMS supports both the 3G-324M and the video over IP approaches, at the server and client. We provide IMS client frameworks, for example for the Samsung handsets that AT&T is selling. A number of the newer Samsung handsets have our IMS software in them. So the idea with video appears to be, if it is person-to-person, the application that people want is two-way audio and one-way video for see what I see.
The other thing that is very big is interactive sessions between you and a server. Either you are watching video clips that you are downloading from a server or you are interacting with the server, choosing the video clip you want to watch.
The question becomes: ‚ÄúHow interactive can you make it?‚Äù I have a Verizon V CAST phone, and it is pathetic, you go through so many clicks to select something, then it takes from 10 to 20 seconds until it actually starts playing. In Europe and Asia we can do a lot better than that. So, that is one application, and it is either for viewing news, or weather, or music videos, or calling a movie theater hotline to order tickets and being able to watch trailers. There are dozens of applications here. People understand it, it is like a phone call, you call a phone number, you interact with something, and you see the video you want.
In the other direction, and we are seeing this in Europe and Asia, people are using their phones as webcams to record things and upload them to an Internet server. The big thing, which unfortunately NMS is not involved in, but some friends are, is a UK service called See Me TV, where they incent people to upload videos by charging them a buck or so for the upload, but pay them a penny every time someone watches their videos.
UK TV news has stories of people making hundreds of pounds because they posted a really popular video. Whether it is things like See Me TV, or dating sites, or social networking, or people wanting to post videos on their personal home pages, the mobile phone is going to be a mobile webcam. Just as camera phones have swept the world, mobile webcams are going to be available to everybody eventually.
I see much more video content being created in the next decade than anybody looking at YouTube today would imagine because everybody will have a portable webcam.
Joe: So that will not only have huge implications for access bandwidth as we discussed earlier but also for storage.
Brough: Video is compute-intensive, mobile video can be done with both the Internet paradigm and the telephone-calling paradigm. There are a lot of ways you can make it blend into what a consumer wants to do, and yet the user interface can be developed independently of the underlying technology.
You have to pick out the things that people actually want, and there is an enormous set of opportunities there, that is our principal focus at the moment,
Joe: I read the other day that a Lucent Bell Labs spin off in Colorado is coming out with a fully holographic data recorder a little later this year.
Brough: In 1998, a magazine asked me to project what would be happening in 2018, 20 years later. The article I wrote was about 3D holographic virtual presence communication, which I projected would become feasible by 2012, at least in the lab, and readily available by 2018. I think I was pessimistic. I think it will be readily available certainly in 2015 and maybe in 2012. Certainly it is beginning to be demonstrable right now in labs. I gave a talk at the IMTF video conferencing conference in Germany two years ago at which I said that people did not want to do video conferencing and that it did not work as well as being there and probably would not until we got 3D holographic virtual presence.
With holographic virtual presence, you get more of a sense that you are talking to a person across the same table and that the person is really in the same room with you.
It might be the thing that video conferencing people have been missing for the last three decades. Certainly video conferencing up through and including the immersive conferencing that people are peddling today is no substitute for being in the same room with somebody.
Joe: Do you see a relatively unfettered regulatory environment in this country and others or do you see increasing government intervention into some of what have become core technologies?
Brough: I see two domains where there is government intervention; one of them is in the right-of-way in front of my house. That is in the end, the justification the government has for regulating the power companies and the phone companies, and the cable companies in their access to limited right-of-way in front of my house. Personally, I would like to own the fiber from my house to some aggregation point where enough other fiber came together that you could get competing ISPs to show up. I would be happy to pay what that fiber would cost. If I did it on a condominium basis with a bunch of my neighbors, it would probably cost me less than $1,000 to $1,200 ‚Äì one time. That is what communities do if a whole community can get together and do it themselves, except that is politically or legally impossible in most places. So that is a real obstacle.
On the other hand, the wireless domain is also improving with Moore‚Äôs Law. There is a long way to go before wireless has the capacity of fiber, but we do have competition. The most encouraging thing I have seen in the United States is that in the Advanced Wireless Services (AWS) auctions last fall, T-Mobile US spent (I think) 3.6 billion dollars buying additional spectrum across the country. They subsequently committed to 2.7 billion dollars to build a radio network using that spectrum, so that means we are going to have four national carriers with advanced 3G networks by the end of 2008. Going into 2009, we will have four different companies with national footprint able to deliver a megabit per second down and several kbps a second up.
Four is a magic number in overcoming oligopoly. Consider mobile operators selling voice services in developing countries. If you have one operator it‚Äôs hopeless; if you have two, you get some mobile phone adoption; if you‚Äôve got three you have growing phone adoption, but if you have four, things go crazy. Four operators is enough competitors to tip the balance. If you have four viable competitors what you get is rampant growth in whatever it is they are competing for. Most of the stuff I have watched is basic voice service in emerging markets, but what I am expecting to see in the U.S. in 2009 is competition in broadband wireless Internet connectivity. With four competitors somebody is going to break ranks and offer flat rate data, that is, up to x gigabytes per day for a flat rate per month.
Consumers want to avoid the risk that they might be charged overages, so they will pay more to get a bundle than it would be to buy the permanent price. People will jump on any flat rate guarantee. I am expecting to have reasonable flat rate Internet connectivity in the U.S. in 2009 on your mobile device. It will still be nothing like the fiber connectivity I get at home, but it is a start.
Joe: RF spectrum: are we going to run out of it, or will we just re-allocate things like VHF TV channels that we are not going to be using anymore?
Brough: The whole thing about spectrum is a big mistake: The idea that there is a limited amount of spectrum is based on the technology we had in the 20th century.
The reason we talk about spectrum is because receiver technology was very limited. If you think about visible light, which after all is part of the electromagnetic spectrum, nobody is regulating anything. There is a white noise source, otherwise known as the Sun, which is streaming incredible amounts of energy in at all times, and yet we pass more information in the visible light spectrum than the entire rest of whatever. With new receiver technology the whole way we approach spectrum is rapidly becoming irrelevant in a technical sense.
Unfortunately we‚Äôve built up a ton of laws, regulations, and vested interests since 1925 when the first federal radio regulation came into effect. I don‚Äôt have any clue how we are going to be able to unwind this, but over the next 20 to 50 years, there will be less and less need for any of the existing regulation. It is a political problem because we created a bunch of laws that should not be there. Worse than that, because we thought the spectrum was limited and we had to allocate it to people, we then went ahead and said we would regulate speech in ways that would otherwise violate the Constitution just because radio waves were so special. So there are just tons of regulations that are based on 20th century technology, not on physics. That is a problem that will take decades to unwind.
In terms of what is going to happen in the next 10 years, there are desperate attempts to claw back bits and pieces of spectrum, which were given out for use in an incredibly inefficient fashion. For example, the television spectrum is being used inefficiently. It was given out based on the quality of the electronics that could be built in the 1950s. The last time the receiver specs were updated was for color television.
Moore‚Äôs Law has come a long way since 1950, but the television bands are still being used in a fashion that is based on 1950s technology. So we will claw that stuff back, and the reason we will succeed in doing that is that Congress is extremely interested in spinning stuff off and getting billions of dollars ‚Äì as with the AWS auctions that happened last fall.
Here is another interesting thing: If you take a spectrum analyzer in downtown Washington D.C. and look across the electromagnetic spectrum from zero to 10 GHz, most of the spectrum is not in use, most of the time. There are a couple little slivers of TV, there are the cellular bands that are widely used, there are a few taxicab slivers, and the entire rest of the spectrum is not in use. There are just a few slivers that have stuff happening in them. So basically we are not using most of the radio spectrum at the moment; we have given out the TV spectrum in a fashion that has such enormous guard bands that in any practical sense most of it is not in any use in most cities.
The most interesting thing that is on the table, which the existing licensees are violently opposed to, is the idea of letting people broadcast in spaces that are not currently in use as long as they have a radio that automatically backs off if anybody else starts to use the space. So even if you have a license there, someone else could use it as long as their radio was designed to a certain spec that guaranteed that if you came on the air with your frequency, that they would get out of the way within two hundred milliseconds or so.
Joe: Or [Ultra Wide Band] UWB where you go underneath everybody.
Brough: Absolutely. The UWB deals with high capacity over short ranges ‚Äì you won‚Äôt need cables in your living room.
With regard to the spectrum issue, the thing that makes sense in terms of the physics would be completely open spectrum and no regulation whatsoever. It may take a hundred years to get there, but that is what the physics would suggest is the ultimate solution. The only band where completely open spectrum and no regulation whatsoever apply at the moment is visible light.
And there are vested interests that have been given exclusive licenses to hunks of spectrum. They are auctioning more and more of it all the time, whenever they can claw something back, Congress auctions it off and makes billions of dollars, which creates more vested interests.
One of my political arguments about broadband access is that the best possible thing would be if the federal government made it so the 22,000 local municipalities could do anything they wanted and could not be blocked by either state or federal lobbyists, because then suddenly you would have competition between local governments.
In an international sense, the good thing about the world today is that you can go to a congressman and say, ‚ÄúLook, we used to be number one, and now we are headed toward number 19 in broadband connectivity. We are falling farther and farther behind.‚Äù When the world gets to the point of dealing with open spectrum issues, that is the kind of argument that will be used, but I cannot guess who is going to get there first.