There is no doubt that inexpensive bandwidth exists these days—and a lot of it. However, it’s not worth much to carriers if they cannot figure out how to use it.
As metro Ethernet picks up momentum as a key access technology, different flavors of passive optical networks (PONs) are gaining in momentum. GPON is becoming more prevalent than the ATM-based PON, as carriers realize that fiber offers superior performance with fewer operational headaches.
That means there are alternatives to using expensive SONET add-drop multiplexers, since carriers can now offer higher bandwidth without truck rolls, forklifts and upgrades.
Gigabit Ethernet offerings no longer have to be dedicated SONET channels into enterprises.
As carriers offer fiber rings to connect multitudes of enterprise customers’ buildings, standards have become the key to differentiating and prioritizing customer traffic riding over metro Ethernet rings.
One of the key enablers for metro Ethernet has been the IEEE 802.1Q standard, which made VLANs possible. Enterprises have long used the standard to virtualize traffic on their own networks. “The beauty of it was its simplicity in making one circuit look like many,” says Jim Brunetti, director of network engineering with Masergy. “For enterprises, they could separate engineering networks from accounting networks from marketing networks. For carriers whose fiber rings serviced multitudes of buildings, it meant differentiating customers riding over their fiber in metro Ethernet rings.”
With 802.1Q, the layer 2 header comprises 12 bits. By adding an extra field in the layer 2 header, 802.1Q enables as many as 4,096 VLANs—each with its own customer ID—over a single circuit.
That extra field possesses a “P” field, which adds another three bits. That enables carriers to do layer-two class-of-service prioritization. In other words,802.1Q enabled differentiation of customer traffic byalloweing service providers to assign a “VLAN tag” or “P-VLAN” to each subscriber’s service via IEEE 802.1Q tagged Ethernet frames.
“But with so many enterprise customers and buildings to connect, the concern has become one of running out of VLANs before they run out of capacity,” says John Holobinko, managing director of Northington Systems, a consultancy specializing in telecom planning and strategy.
As more operators offer metro Ethernet for access, more VLANs will be needed to maximize the use of ever-expanding bandwidth. With talk of 10, 40 and 100 Gbps metro Ethernet offerings, 4,096 VLANs won’t be enough when that comes to fruition. Of course, that is still some way off, as most offerings are still in the 100 Mbps range. “In the old days, 4,096 VLANs over a 1 Gig LAN was huge, but more VLANs become necessary if the bandwidth on these pipes continues to grow,” says Holobinko.
With 802.1P, there was the addition of three more bits, which added an extra field in the layer 2 header, but even that is not enough to make the best use of a 10 Gbps metro Ethernet offering. Although the “P” bits enabled carriers to assign layer 2 class-of-service designations, more was needed.
As a result IEEE 802.1 Q-in-Q has evolved over the past few years as a key technology in Ethernet services. Also known as “VLAN stacking,” Q-in-Q means carriers can scale services and increase the number of VLANs at their disposal exponentially by adding another header to the front of packets. “Where there were 4,096 ID possibilities with one header, the number is squared with the addition of another header,” says Brunetti, explaining that switches act initially just on the outer header. “A pipe possessing 4,096 tunnels between cities will gain an additional 4,096 tunnels within each of the original tunnels. Switches focus only on the outer header until traffic gets to its destination city, at which time the header is stripped off. As traffic goes to each customer [identified by the VLAN tag], the inner header is stripped and the customer can see its Ethernet packet.”
Even though these evolutionary steps have allowed carriers to grow their networks and classify services with ever greater efficiency, a lot of work remains to be done.
“We still have to wait for carriers that we use for our local loops to really embrace metro Ethernet and all of these facilitating standards,” says Brunetti. He notes that a lot more fiber is in the ground in Europe and Asia. “While some smaller players have a few lit buildings, and several RBOCs are starting to make announcements, it’s still not going to be widespread and available as a wholesale offering until the end of ’06,” he predicts.
C-level executives of the major RBOCs all have made metro Ethernet announcements, but very few real products are really mature. The manufacturers are still scurrying to address interoperability issues, as well as those around operations and management. “There’s a lot to flesh out still,” Brunetti says, noting that the strict rules engineers have come to know in TDM do not exist in Ethernet. “With TDM, there were strict rules for finding errors, identifying devices and fixing them. With Ethernet, the only thing that is really known is the connection to the next switch; there is no end-to-end visibility for troubleshooting and management yet.” As those issues are resolved, he says, more carriers will forge ahead with metro Ethernet.
Email this article
Add a comment
Printer version
Order reprints
RSS Feed
Bookmark article