 |
 |
 |
 |
 |
 |
 |
|
|
|
||||||
|
||||||||
|
||||||||
 |
||||||||||||||||||
|
||||||||||||||||||
|
NGN
- The Future of Networking
Next Generation Networks is a first class conference crafted by John McQuillan. In a nut shell the conference reflects the struggle between the old and new with the old being the traditional phone connection based network and the new being a packet based connectionless network. But it is much deeper than this. It is also about electronics verses optics with the latter achieving 2X performance every 9 months while electronics based on Moore's law lags with improvements on an 18-month basis. Thus, from an implementation perspective network architectures are seeking to push electronics to the edge of the networks as optics increasingly dominate. However, optics has severe limitations in that active circuits are not available and the central function is to switch. As a result optics at a transport level is point to point, or at least, lambda grained with WDM fiber, it is point to point. Conference Overview NGN is being held in a hotel across from the Washington, DC Convention center. The hotel is strained to capacity to hold the event. The tradeshow floor is tiny and cramped to the point where not another booth can be accommodated. The isles are packed such that it is nearly impossible to walk by many booths. Yet, the conference chairman, John McQuillan, announced that it will move to Boston, but it will not become a trade show. He feels it essential that for the personal nature of the show to be preserved the attendance must be capped and not allowed to grow unchecked. For this event he stated that 100's were turned away. This show ranks in networking with the Seybold events - first class. Every attendee gets 3 volumes of all the talks, which must be submitted by the speakers 2 months in advance so that all the books can be printed. Further, there is a CD with all the talks on it. As one of the speakers stated - John runs the show with military precision. One of the problems is that there are too many sessions with high interest topics. A session on moving fiber to the metro network was so packed that they were passing out chits so those waiting could enter in the proper order when others exited. In spite of these limitations we picked up only positive comments on the show. Attendees were also present from all over the world. Intelligent Optical Networking Where are future networks going? Part of this was addressed in an excellent tutorial given on Optical Networking Technologies. Today, there is an overlay of two optical networking technologies taking place at the same time. The more traditional solution based on a ring topology is SONET. However, this has major limitations. Some of these include interfacing with other SONET rings and transport of non-ATM protocols. The favored technology is mesh, which is a more traditional networking approach. However, the implementation of this is very much based on optical technologies and this is the basis for Intelligent Optical Networks. The layering of the Intelligent Optical Network, from the bottom is: fiber, wave, lightpath and service. A key point is that the transport is based on lambda - the single frequency band used for transport in a WDM fiber. It is the add/drop and switching of these frequency bands which provides the basis for optical networking. It is here that switching takes place but this is quite crude - compared to packet switching. One of the reasons for the crudeness is the limitations of optical components and the lack of visibility, with optics, into the contents of the stream. Yet, these limitations should not be seen as the foundation for significant drawbacks of optical networking. The catch phrase is - if it is bits use electronics, if it is streams us optics. What Intelligent Optical Networking provides is an end-to-end delivery of ATM, IP, Ethernet and OC-N services. Note that these are both connection based and connectionless. In fact, the driver for many of the new services is GE (gigabit Ethernet) and 10GE. A point made many times in the Intelligent Optical Networking presentation is that the network can be "point and configure" networking even at the user level. The most significant factor in this technology is that with the transition from SONET based networks, which are the purview of the provider, with Intelligent Optical Networking the network and its configuration can extend directly to the end user. Thus, there is significant value added to the network when the customer is a part of the optical interconnect and service definition. Where the Net Is Going Next - John McQuillian John is a deep thinker and has an uncanny ability to weave his observations into a very compelling presentation. He did this again at NGN2000. To address the topic of his talk he examined the contradictions surrounding the status and future of the net. To accomplish this he presented a thesis then an antithesis and a synthesis of these two views. In this way he could expose the many contradictory views of what is happening in the evolution of the network. The cute part of the presentation was his ability to inject the views of the profound thinkers of communism - interpreted in a next generation network framework. An excellent example of this approach was in addressing the future of the network - a thesis would be - the future of the net looks like the present but it will be bigger, faster and more of it. The antithesis is that there is a revolution underway which is expressed as NG providers will build NG networks using NG technology from NG vendors using NG funding. Yet, a view which represents the synthesis of these is one of organic growth where the old and new will coexist in a complex whole. In this context he then proceeded to analyze the NGN future for industry structure, capital market dynamics, network economics, key bottlenecks, network architecture, dominant culture and service provider ecology. It was not only entertaining but deep in its perspective on the forces which are shaping the development and evolution of a digital infrastructure. A recurrent theme in many of the sessions is the contradictory messages being sent by Wall Street on the future of telecommunications. In recent months most of the stocks of both the large telecommunications companies have been battered and the staples of the future, such as Cisco, have been hammered down. An interesting view presented by John McQuillian is that the intensity of M&A, as a means of liquidity, will continue in spite of the downturn. He stated that in the service provider space the advantage goes to the incumbents with consolidation for the new players being the norm. While in the network chip space incumbents will acquire nearly all the startups. In the system vendor space startups will challenge the existing companies, including Cisco and Lucent, to the point where these companies will find it difficult to continue to grow. For the major carriers a bet-your-business crunch lies ahead which centers on capital expenditure in the transition to the new network architectures. In 1996 90% of the revenue was from POTS while the capital expenditure was $43b based on a $218b revenue stream. In 2000 2/3 of the revenue was from POTS but the capital expenditure will be at 3:1 of the revenue - $326b of revenue and $105b in spending. While in 2001 the spending on cap ex will be 1/2 of the revenue stream. The crunch comes in 2005 when only 1/2 of the revenue will come from POTS. Only those carriers who have invested heavily in infrastructure for non-POTS service will survive but this comes with a very heavy toll of investing. Yet, the contradiction in today's stock market is that the value of the communications companies has plummeted. As John Sidgmore, Vice Chairman of WorldCom stated later in his keynote - many of the telecommunications companies are trading at book value. With that it is easier to divest rather than stay in business as a whole. One of the best illustrations of the transitions taking place in network lies between SONET and Gigabit Ethernet. There is a widely held view that SONET, the original fiber based implementation largely based on ATM, is legacy in the context of the need for Internet based services. John McQuillian provided interesting statistics on both. SONET spending was $7.3b in 1999. WDM is growing at 98% per year and SONET is only at 62%. There were 2.3m SONET and SDH modules shipped in 1999. Gigabit Ethernet is to grow 200% in 2000 to $4b. A 10 Gigabit Ethernet port is expected to cost $3,500 while an OC-48 costs $50,000 and OC-192 is $180,000. CIBC predicts that Ethernet will appear as a public service by 2004 and the value of that market will be $15b. Venture Capital Investment Roundup Venture investors have been rocked by the turmoil in both the dot com sector and telecommunications. Geoff Yang, of Redpoint Ventures, asked probing questions about the continued high evaluations of the telecommunications equipment providers while the buyers, the telecommunications companies are in the tank. He articulated a key point which underlies the drive to optical networks - while Moore's law is laggard with improvements in performance on a 18 month cycle optical performance doubles every 9 months. As a result, there is great pressure to eliminate as much silicon from the network as possible. The underlying reason for this is that in order to keep up with demand only optics is viable. Thus, electronics is getting pushed farther and farther to the edge of the network. Geoff also stated the hot areas for investments include: optical components, Internet data center, and storage service providers. He concluded by stating that the landscaping changing events are Ethernet and wireless. Iospan Wireless Tucked away in a session on Fixed Broadband Wireless Technologies was a presentation by Arogyaswami Paulraj, Founder, Chairman and CTO of Iospan Wireless, formally known as Gigabit Wireless. Paulraj is a professor on leave from Stanford while he helped form this company. NGN2000 was the coming out event for the company. For the first time they described the challenge of BWA (Broadband Wireless Access). At the same time the company released a press announcement on the company, its technology roll out and the name change. The requirements of the market, which appears to be largely based on Sprint's buy requirements for its MMDS buildout, are: -Symmetric speeds
> 4 - 5 Mb/s Chart after chart he described how difficult this problem is. In the end the following is required: Multiple Input Multiple Output (MIMO) antenna technology where there are multiple transmit and receive antennas. This is seen as critical to accomplish the requirements above, especially in hostile RF environments in the MMDS bands, where Non-Line of Site (LOS), communication is used. MIMO allows for spatial, polarization and temporal diversity. ODFM modulation
The following tidbits came from the talk, questions and booth discussions. Iospan has 60 PhD's working on this. Iospan has completed its own outdoor tests and is now integrating its products. A cell is based on 3 sectors and thus there is frequency reuse. There was a prototype base station receiver and consumer antenna in the Iospan booth. However, it was not clear that these were populated equipment samples. The CPE antenna has 4 discrete antennas in one housing which is approximately 18" X 9" X 4". Given the size it is not clear how this can be a consumer self install, especially on a window. First generation implementation is based on a base station line card which services a 2 MHz band that is translated into a 13.5Mb/s downstream and 6.75Mb/s upstream path for that sector. The back haul bandwidth uses OC-3. A single cell site is capable of supporting 15,000 users but the bandwidth is shared by these users and this is in turn determined by the spectrum available for that cell. The operational model for the first generation systems is a cable modem - many users sharing limited service capacity with the emphasis on the downstream bandwidth. Thus, this model is well suited for Internet cruising or streaming media downloads. Although the system can certainly support video to the home including video conferencing the most valuable commodity is spectrum and this limits what can be done. This is the basis for the cable modem deployment model - spectrum makes bandwidth a shared resource among all the users on in a given time. Units will ship for evaluation in Q2 2001 of which Iospan expects the first units to go to Sprint for field trials. Iospan feels it can get to $200 to $300 CPE but not on the first generation. Sprint has delayed its original purchase of equipment for the implementation of broad deployment of MMDS services, which was originally scheduled to take place in Q3 2000. The WAVE Report believes this to be due to the poor performance of current solutions which are not as sophisticated as those provided by Iospan. Further, Sprint has set very aggressive schedules for component prices which others have stated will be hard to achieve. Iospan feels that it just takes time to get the technology to work which we regard as another way to state it feels it has a competitive opportunity in the Sprint buy and Sprint feels it has to delay the buy. Cisco is a player in the MMDS market because of its purchase of Clarity. Some of the employees of what was Clarity were also students of Professor Paulraj. Cisco is apparently adopting MIMO technology in order to get adequate performance and to be able to compete in the market, i.e., the upcoming bids for equipment buys by Sprint and WorldCom. |
|||||||||||||||||
| Comments? |
||||||||||||||||||
