Who guarantees your “committed” Ethernet traffic?

Yossi Saad — Fri, 10 Jul 2009 15:38:34 +0000

Many applications require high quality of service (QoS) transmission to make the service operate properly. Audio and video streams suffer from loss of packets that translate into audible blips and picture frame loss, reducing the end user’s quality of experience (QoE). Ethernet is a ubiquitous, cost effective technology that enables delivery of multiple services with proven efficiency and simplicity. However, with carrier Ethernet (CE) proliferating into the transport network and replace the legacy SONET/SDH networks, we should take a deeper look into it, making sure it can live up to its promise.
Ethernet technology has evolves from its enterprise days to cater for the needs of carriers. The VLAN tags used in the enterprise to separate between departments have been enhanced with “double tags” defined in the Q-in-Q (IEEE 802.1ad) standard. This step was perceived by many as a step that was sufficient to make CE a “carrier-grade” technology. A further enhancement aimed at increasing the scalability of the solution was added (PBB, 802.1ah), and the migration to packet networks seemed smoother than ever… or so it would seem.
Classical Ethernet terminology is misleading, at least for transport engineers. Whether a Q-in-Q or a PBB technology is used, Ethernet services use a parameter called: “Committed Information Rate” (CIR). This CIR value is used to mark packets that are sent through and are not exceeding the CIR rate as Green frames, meaning they get priority in the network queues over excess traffic of the same class. However, when such a term is used in the transport market, it is translated to an implied message that the traffic flow is indeed guaranteed. The service provider and its customers think they can rest assured that all packets will get to their destination just like in the old SONET/SDH days, right? – Dead wrong!!!
The use of CIR to mark Green frames is good to ensure priority of the “committed” traffic (for which the customer paid a lot) over the “excess” traffic (i.e. the extra bandwidth for which the customer paid less). However when we look at the traffic flow in the network, “committed” traffic from many users can concentrate on a specific network resource, e.g. a link, and exceed the capacity of this resource. In such case, some frames must be dropped, meaning losing part of the so called “committed” traffic.

For example, in the above figure, assuming the traffic from all 3 customers flows into switch F through switch D, the link between switches D & F becomes congested and therefore over 10% of the “committed” traffic will be dropped. To avoid this congestion, the traffic from switch C could be routed to go directly from switch E to switch F, but unfortunately the service provider has no control over it. The routes are determined automatically by the spanning tree protocol (STP) and could change due to network rearrangements without service provider’s involvement.

Carrier Ethernet 2.0 has evolved to overcome this issue, and make CE as a real alternative to SONET/SDH. It has introduced the concept of “connection-oriented” Ethernet (CoE), so that the service provider can have full control on the paths taken by the services and reserve bandwidth for committed services. CoE has been recently standardized and ratified by IEEE as 802.1Qay, a.k.a. PBB-TE. With CoE, the control plane functionality is implemented by an external management system, provisioning the trunks each service must follow and ensuring bandwidth allocation meets the requirements. There could not be any contention between committed traffic since the management system verifies there is sufficient bandwidth for all committed traffic throughout the whole path before the service is configured and approved.
Carrier Ethernet is now ready for prime time. The combination of connectionless Ethernet flexibility and connection-oriented service guarantee, combined with connectivity fault management for service assurance deliver on their promise. Now there is a viable, cost-effective replacement for legacy TDM.

How would Carrier Ethernet succeed where ATM failed?

Yossi Saad — Thu, 09 Jul 2009 10:34:12 +0000

Carrier Ethernet (CE) is a hot topic for the last several years. We have seen a lot of media hype, significant technical work done in industry forums and standards committees, and marketing activities sponsored by various industry players. Originally named Metro Ethernet, the repositioning of the technology to support the whole carrier network changed its name to be Carrier Ethernet. It was expected that CE would be replacing the legacy infrastructure of carriers, with the promise of an all-packet, unified transport solution, delivering all services for all purposes.
But hey, doesn’t that sound familiar? If we replace in the last sentence the “CE” for ATM and the “packet” for “cell”, wouldn’t it be exactly like the message ATM supporters conveyed back in the 1990s? ATM took a leading position in core data networks and thanks to the ADSL-standard adopting it, in the access DSLAMs, but it had never been able to overcome the complexity and cost barriers and become the single, unified transports its proponents so desired. So if ATM could not make it with all the efforts and money spent on it, why do we feel so comfortable that CE could do it?
I think the difference lies in the two critical factors mentioned above, cost and complexity (maybe we should brand these as CO2…). My years of experience in this industry have taught me that these are two of the most critical factors in any decision making process in the service providers’ market. ATM was simply too complex to carry multiple services, compared to the simple SONET alternative. Since at that time most of the traffic (including data) was still based on the T-carrier system, SONET was much more cost-optimized and simple compared to the adaptation layer required for ATM transport. Yes, timing is everything…
Now let’s apply the above lessons to our current environment. We all recognize the growth in data services that are based on Ethernet/IP. Legacy TDM services are declining and are headed towards extinction (though it will take quite a few years of course). The data bandwidth far exceeds the real “TDM” traffic, SONET/SDH can no longer meet the demand and need for a packet-based network is evident. However, as for the choice of the right packet technology, we need to turn back to the textbooks and read the chapter about CO2 (I knew this phrase would catch up). We have two relevant technologies to choose from: Ethernet, ubiquitously deployed in the enterprise and growing in popularity in the provider access networks, and IP/MPLS that is often found in core networks. Service providers have now two options to choose from: Carrier Ethernet, natively simple and cost effective, or MPLS, already existing in their network core and promised to be cost reduced and somewhat simplified to enable widespread proliferation.
At the end of the day, as coined by Bill Clinton in the previous millennium – “it’s the economy, stupid”. The cost is the key factor, especially in our current economical downturn. Complexity is also translated to cost – training existing technical staff, recruiting new staff, spending time and effort due to long learning curves etc. Ethernet provides a solution that is similar to today’s data access networks, and with the appropriate management approach, also resemble legacy SONET/SDH. It is surely going to play an important role in the transition of our networks to the all-packet dream.

Telecom Networking

Who guarantees your “committed” Ethernet traffic?

How would Carrier Ethernet succeed where ATM failed?