Ethernet as a Carrier Service (Cont'd)

So clearly network service providers had to start delivering Ethernet to their enterprise customers, and many of them are doing so, as described elsewhere in this edition of Pipeline.

When carriers started to think about deploying Ethernet in the wider network, LAN folks everywhere approved. Their familiar, forgiving and inexpensive friend would escape the walled garden of the LAN and get to play in the big world of the WAN. LAN people would no longer need to understand arcane technologies like SDH/SONET, Frame Relay and ATM. With Ethernet everywhere, and IP running over it, the global network would be elegantly simple and easy to manage. There would be one unifying Layer 2 technology spanning the globe.

At the same time, telco folks allowed themselves some skepticism. It’s their job to be skeptical about the deployment of new technologies. They have a tradition of delivering their customers the highest possible standards of quality, reliability and security. Could a technology conceived and designed for the LAN environment possibly make it as a WAN technology? Ethernet needed to meet the WAN challenges of QoS, reliability, and security while being able to handle WAN volumes of traffic, and WAN geographical distances. Clearly, these are not trivial challenges. Nevertheless, driven by the prospect of even larger global markets for Ethernet technology, the network technology community has come up with some answers.

Ethernet uses Carrier Sense Multiple Access with Collision Detection (CSMA/CD). Isn’t this technique inherently unscalable to the sort of traffic levels used on carrier networks?
Apparently not. Ethernet networks at any level need to be designed with sufficient overhead to minimize congestion. However, higher bit rates and the ubiquitous use of Ethernet switches reduces the number of simultaneous messages on any one segment, and higher bit rates throughout enable more transmissions to take place without collisions. It’s scalable.

Ethernet is intrinsically a short-distance protocol. LAN distances are trivial, but telcos need to think about distances in transcontinental or even global terms. How can Ethernet be adapted to long distances and still be Ethernet?
The distance challenge is not related to the Layer 2 data link design of Ethernet, but to the various choices that had been made for Ethernet Layer 1 physical connectivity, which traditionally looked for pragmatic solutions for LANs. However pragmatic solutions for metro-Ethernet and long-haul networks exist too, now that we’ve started putting them together. In the WAN arena we can now use 1000BaseLX or LN GBICs (Gigabit Interface Cards) over certain legacy 1310nm single mode fiber to reach distances of up to 10km, further with regeneration. (The approach is similar to the LAN standard 1000BaseSX, in that it sends and receives Ethernet directly over dark fiber, but SX is distance-limited because it uses cheaper multi-mode fiber.) New implementations can benefit from new generations of fiber, for example 1000BaseLX can be made to work over 70km or more using 1510nm NZ-DSF single mode fiber. DWDM (Dense Wave Division Multiplexing) techniques can take us even further: Gig-E or even 10Gig-E can be transmitted up to 500km before regeneration is necessary. All of these developments provide a range of technologies suitable for deploying Ethernet at metro, regional, and national levels.

« Previous   |   1   |   2   |   3   |   4   |   Subscribe

Send Comment

© 2005, All information contained herein is the sole property of Pipeline Publishing, LLC. Pipeline Publishing LLC reserves all rights and privileges regarding the use of this information. Any unauthorized use, such as copying, modifying, or reprinting, will be prosecuted under the fullest extent under the governing law.