A Practical Navigator for the Internet Economy

GIGABIT AND TEN GIG ETHERNET BY PASSING CARRIERS AND REVOLUTIONIZING TRANSPORT AT INTERNET'S EDGES

GIGABIT ETHERNET COMBINED WITH METROPOLITAN AREA FIBER ENABLES NEW DATA NETWORK BUSINESS MODEL IN MONTREAL,

pp. 1-9

We interviewed Francios Menard, Lead Engineer Internet Division of Cable VDN in Montreal. Cable VDN is a Montreal competitive cable TV provider. Menard is designing and implementing a gigabit per second Ethernet data network over the same fiber infrastructure by means of which it delivers cable TV. As far as access to fiber optic cable and a city wide infrastructure of conduits goes, Montreal and Stockholm are possibly the most fiber rich cities in North America and Europe. In Montreal VDN owns two cables, each extending for more than 200 kilometers beneath the city. One cable has 288 strands and the other 144 giving it well over 10,000 kilometers of fiber through out the city.

What we see with VDN in Montreal is the beginning of new business models for start up data network service providers. For customers who do not need to interface their data network with a carrier SONET based network, it is possible to build a gigabit per second data net using gigabit Ethernet as the framing and transport layer. The cost of doing this is extraordinarily cheap compared to carrier SONET based managed data services. This leads to five year contracts for point to point, kilometer long, gigabit per second data services at a thousand Canadian dollars per kilometer per month or less than $700 per month U.S. dollars. The cost then for a thousand megabit per second Ethernet data network service looks to be about 20% more than the cost of fifty megabit managed SONET data services from a carrier. Twenty times the service for only one fifth more money. This price differential makes it possible to "over provision" most customers who when their data use peaks will have a much higher ceiling of bandwidth on their data lines. If their traffic bursts never hit their connections upper boundary, congestion is unlikely to develop making those pesky quality of service issues on their networks irrelevant.

Of course bandwidth demand will likely grow to meet whatever amount is available. Nevertheless Menard suggests that Gigabit and soon ten gigabit Ethernet data service makes it possible to cease emulating the PSTN with protocols like MPLS serving to route restricted data services over permanent virtual circuits where network applications are designed with the expectation that they will be bandwidth limited. According to Menard gigabit Ethernet makes it possible to design the network with from the point of view of unlimited bandwidth and applications that "adapt" to the bandwidth available to them. For example audio codecs can be used that a very stingy way if the network is crowded. However the user can shift to mpeg2 and then to mpeg3 when bandwidth is plentiful. He suggests that over provisioning with gigabit Ethernet will give application designers room to experiment in creative ways with new applications which will become attractive precisely because they are adaptive. The large initial amount of bandwidth should be attractive to network MIS managers who won't have to be measuring their network consumption to order new lines on a monthly basis.

As adaptive applications are be designed they must incorporate means of locking out users who refuse to throttle back to lower levels of bandwidth consumption when signs of network congestion appear. The network manager will have to set policy for over all application use since some applications can in the hands of some people suck up most available bandwidth. In a well designed network where application tools have been correctly provisioned such tools should be able to talk to the network management system and smooth out usage patterns. What is driving this remarkable growth in bandwidth is the ability for the first time to go SONET free in a high speed network and the dramatic increase in the speed of Ethernet which is the most thoroughly comoditized and wide spread technology in data networking and therefore the cheapest means of data transport.

10 GIGABIT ETHERNET DRAFT ALMOST OUT GIG & 10 GIG E ENABLE SONET FREE IP DATA NETS

AVAILABILITY OF CHEAP TRANSPORT TECHNOLOGY FOR BOTH MAN AND WAN ENABLES ENTERPRISES TO DROP CARRIER MANAGED SERVICES,

pp. 10-15

We interview Cisco Distinguished Engineer Howard Frazier. Frazier is one of the inventors of ten gigabit Ethernet and a key technical contributor to the standards activity. The draft ten gigabit Ethernet standard is due for delivery next month (September 2000). They have selected two specifications for single mode fiber. One spec (the LAN) is good for cabling up to ten kilometers on single mode fiber. The LAN spec is intended primarily for campus applications and for connection into DWDM transponders. The same spec will be good for hooking a switch or a router to a DWDM box. Ethernet speed here is defined as 10 gigabits per second.

They also selected a specification for going long distances on single mode fiber. These are distances of greater than or equal to 50 kilometers. In the WAN specification data is slowed slightly. It runs at the SONET OC-192 rate which is 9.29419 gigabits per second. They only specification they have not decided on is the one for multi-mode fiber. They have five or six competing specifications that they must narrow down to a smaller number. Cisco has publicly announced pre standard 10 gigabit Ethernet products including dual port cards for its Catalyst 6500 switch.

The Frazier interview validates what we learned from the study of the VDN situation in Montreal. With an abundance of fiber and Gig and soon 10 Gig Ethernet from a data network point of view, the different between MAN and WAN (Wide Area networks) has disappeared. Frazier gives examples of gigabit Ethernet service on leased dark fiber from New York City to Princeton New Jersey, a distance of about 100 kilometers or sixty miles. Fortune 500 companies with offices in New York have bought managed SONET data network services from the carriers in droves. Fifty megabit guaranteed service costs $16,000 a month from the carrier.

If the same company has leased dark fiber from Manhattan to Princeton, it can get gigabit service (twenty times as much of what it is giving up) to Princeton with no carrier involved. It can do this for the $15,000 cost of a pair of one gig Ethernet port cards for its enterprise switch. The only other cost is staff and fiber leasing experience. After the first month 20 times the telco services become free because the company uses precisely this equipment to provide its own service campus to campus. It seems likely that the Fortune 500 will be dropping their carrier SONET based services in droves and doing all this gigabit networking for themselves as their managed carrier data network service contracts mature. Frazier has observed several cases. With gigabit Ethernet enterprise networks have a chance to become SONET free. SONET free means much less cost and more effective use of fiber.

The implications of what is in effect another revolution are profound. The only reason enterprise networks would need to use SONET is if they needed to connect to a carrier network that did. Virtually all do. Now cross country long hauls used to require SONET. They don't anymore. Frazier shows how gig and ten gig Ethernet may now be used cross country. In effect data networking doesn't need the carriers any longer. They will soon sit with hundreds of billions of dollars of SONET switches to amortize and the only thing they will find these switches good for is supporting their long distance voice networks that are no longer profitable. We find that the carriers are going to realize that they need to get their SONET off their books as quickly as possible. One way to do that would be to sell off their voice long distance services.

AN ERICSSON PERSPECTIVE ON GLOBAL WIRELESS STANDARDS AND ON CHINESE WIRELESS MARKET

CREDITS BROAD ASIAN EUROPEAN MARKET PENETRATION TO GOVERNMENT PRESSURE FOR REGIONAL STANDARDS AND INTEROPERABILITY,

pp. 16-21

We interview Gary Pinkham, Vice President of Business Development at Ericsson. Pinkham explains third generation wireless services and their migration paths on a global basis. Services covered are GSM, CDMA, cdmaOne, CDMA 2000, TDMA, Edge, and wide-band CDMA which he thinks will become globally dominant. We asked Pinkham why wireless penetration was greater than in the US. He replied: "in Asia and Western Europe, they have a good path, because they have a good spectrum plan. They allocate the spectrum and specify technology to be deployed. That way they ensure interoperability and a mass market." The U.S. free and open market approach he finds to have left us with patchwork and fragmented services. He compares CDMA and TDMA and Ericsson's relationship with Qualcomm. He finds China which he had just visited to be predominantly a GSM market. China Mobile which is signing up two million new users per month he estimates may grow to 250 million users within five years. China may never install ubiquitous wireline services but leap frog instead into universal wireless.

COMPLEX ISSUES IN IPV6 ADDRESSING ALLOCATION AND ADMINISTRATION

pp. 15, 21-22, 24

From the IETF mail list a discussion of the need for charging fees for the allocation of IPv6 blocs. We find that even though there are several orders of magnitude more IPv6 than IPv4 numbers hierarchical routing will still be necessary.

SOURCE OF STATEMENTS ON JAN. 2000 IETF - ITU MEETING, p. 15

This meeting focused on IP telephony and marked the beginning of IETF and ITU cooperation on Enum and other protocols.