When packet transport first became necessary, the industry's focus was fixated on how to carry packets over the existing non-packet transport infrastructure. Today, the focus is opposite--how to transport all types of data traffic over a packet infrastructure. Ethernet has emerged as the protocol for end-to-end transport across carrier access, core and regional networks.
However, with the rapid changeover to a packet transport infrastructure, some features that carriers were accustomed to were initially absent. Enter carrier Ethernet, which seeks to upgrade Ethernet with features that carriers not only want but also depend upon to execute their business. What makes carrier Ethernet unique, what are the next issues to be tackled, and how are organizations such as the Ethernet Alliance helping smooth the road to adoption?
Mapping the challenge
It might appear to be a gross oversimplification, but here's the best way of distinguishing carrier Ethernet from traditional: with plain, old, vanilla Ethernet, you transport data; in carrier Ethernet, you carry someone else's data.
Imagine you live in San Francisco and your mother lives in New York. Now, Mom wants to send you some money, so she visits her local bank and makes a deposit into your account. You then make a withdrawal from a San Francisco ATM. Of course, the money you receive isn't the same physical cash that your mom deposited back in New York, but this doesn't matter--it spends the same.
Say, however, Mom wants to send you her famous, homemade snickerdoodles. What if she donates the cookies to a New York food bank and then you pick up some others at a San Francisco food bank? Sure, you still get cookies, but they aren't Mom's. The better option here is for your mom to ship her specific cookies via a national parcel carrier. This way, you get the cookies you wanted.
A traditional Ethernet local area network (LAN) works like the bank in the first scenario. All the incoming packets are buffered and sent on their way. It works fine when all the packets are of one entity. However, things become problematic when the packets are disparate traffic of different subscribers; mixing the packets causes multiple problems--security being only one.
Carrier Ethernet strives to work like the parcel carrier with the special snicker-doodles in the second scenario. One subscriber's particular packets are packaged and transported undisturbed from one location to another. Carrier Ethernet augments traditional Ethernet with the evolving capabilities carriers need to successfully transport packets among their customers.
Standards milestones
Carriers' first attempt at packet transport was simply connecting their customers' LANs into a common, shared LAN infrastructure. This didn't work. Multiple customer LANs could have identical Internet Protocol (IP) addresses because of incorrect extension of IPv4, or identical Media Access Control (MAC) addresses because of the use of illegally cloned Network Interface Controllers (NICs). In these cases, there was no way for a carrier router to decipher where to send which packets.
Quickly, it became apparent that there was work to be done to ensure that the various Ethernet streams could be identified and segregated as they were transported through metro and regional networks. Standards work commenced--first at a crawl, then a walk and now a full-speed sprint.
Early accomplishments were designed to rapidly bandage urgent problems. IEEE Std 802.1ahTM-2008 MAC-in-MAC, for example, was conceived to overcome Ethernet's traditional flat addressing scheme. It defined hierarchical MAC addressing to Ethernet that would allow a carrier to have its own addressing, irrespective of its customers' addressing.
Similarly, virtual LAN (VLAN) tagging was designed to enable carriers to correctly separate and transport packets emanating from different LANs. Traditional Ethernet uses a 12-bit VLAN field, yielding 4,095 VLAN tags. That's sufficient for traditional enterprise LANs but not nearly enough for carriers. With the onslaught of service innovations such as triple play, high-definition television, video-on-demand, etc., carriers now must not only differentiate among customers, but also among each of their services. Carrier Ethernet solves this quandary by enabling double tagging, also known as Q-in-Q, where packets contain both outer and inner VLAN identifiers, thus expanding the possible number of VLAN tags.
Twists and turns along the road
Carrier Ethernet needs are not limited to new packet formats only. Today, there is an explosion of carrier Ethernet protocols and standards as the world has steadily migrated to an architecture of IP over Generalized Multi-Protocol Label Switching (GMPLS).
Given that carriers are faced with trillions of packets emanating from millions of customers, bandwidth management is an important area of work. Simply marking packets with priority is not enough, and may result in an important packet waiting in line behind a jumbo packet in the scheduler. Hierarchical Quality of Service (H-QoS) has emerged as an indispensable algorithm for carrier switching engines. H-QoS helps a carrier deliver the expected level of service for latency-sensitive traffic when faced with a deluge of non-critical traffic. One example is separating the voice from data packets when a consumer is simultaneously calling someone and browsing the Internet with a smart phone.
One of the latest trends in carrier Ethernet is work on Ethernet-bridging extensions, which extends early simple bridging protocols to allow data-center virtualization. This is an increasingly important carrier concern with the developing opportunity of hosted cloud services.
Inter-carrier exchanges have now become reality. In a breath of fresh air that is sure to accelerate Ethernet's already meteoric rise, a carrier no longer must own the entire network between two customers in order to guarantee carrier Ethernet performance. At inter-carrier exchanges, carriers may hand off traffic to each other to help further their reach and geographic penetration, while still maintaining packet performance and compliance with service-level agreements (SLAs).
Looking toward the future, one may envision bandwidth-trading markets, as, once you standardize how to carry someone's packets, you can start trading, bartering and exchanging the traffic in an open market. The future of networking might look like an eBay or Chicago Board of Exchange for Ethernet packets--with routes, priority, types, etc., all factoring into a packet transport value that is bid and asked for.
Ethernet has evolved over years from strictly a LAN protocol to the wide-ranging protocol of choice for carriers in their GMPLS backbones, metro Dense Wavelength Division Multiplexing (DWDM) networks and even local access loops. This shift is evidenced in the Ethernet Alliance. Well known and established among enterprises after driving key enhancements for 40/100GE, data centers, 10GBASE-T and energy efficiency, the Ethernet Alliance has extended its focus to encompass the increasingly important carrier Ethernet space, as well. Today, the drive toward carrier Ethernet continues, with scalability that is more efficient and lower latency among the ongoing frontiers of development that will encourage the protocol's continued success and expansion.
Jim Theodoras, Chair of the Carrier Ethernet Subcommittee, Ethernet Alliance, is a monthly columnist for FierceTelecom.
Related articles:
Stimulating carrier Ethernet sales
Ethernet exchange: Another tool in the Ethernet service toolkit