The CBX event in New York last week harked back to the heydays of telecom, while at the same time pointing out the future – there was a buzz in the air as the busy show floor turned into a dynamic meeting place between service providers large and small, working out new wholesaling arrangements to gain market access and deliver new services. That part wasn’t new, the Telx-sponsored event has always been about making deals. But two new trends rang loud in conversations and the conference: the arrival of Ethernet wholesale, and the need for ultra-low latency circuits. Of course they’re related: as critical applications demand ever more bandwidth, they are driving the need for cost-efficient Ethernet connectivity, but not at the expense of performance.
In a way, this situation underpins the need for Quality of Service (QoS) service level agreements (SLAs). Justifiably, SLAs are even more important with a non-deterministic technology like Ethernet than in TDM days past. But conversations with leading operators hinted that SLAs alone might not be enough to secure Ethernet wholesale contracts. Carriers are used to having access to a variety of routes with a range of quality guarantees, switching between them dynamically using least-cost routing schemes. There was certainly speculation that similar systems may soon be needed in wholesale Ethernet.
It’s not without precedent. Least cost routing followed a similar progression in the voice world. Circuit-switched voice was the original application driving the need for routing platforms, sophisticated systems that compare the current tariffs offered by dozens of carriers in real-time, dynamically sending traffic over the lowest-cost routes as capacity permits. When VoIP arrived, a new wrinkle was introduced. Cost and capacity were no longer the only variables: toll-quality QoS was no longer a given. Different encoding and compression schemes impacted calls, so did changes in network traffic resulting in congestion-based packet loss. So least-cost routing systems evolved to factor in QoS as a new variable – test heads now perform thousands of test calls every hour over available routes, allowing carriers to match the lowest-cost route meeting a pre-defined quality to particular customers or applications (e.g. business vs. residential or cellular calls).
There’s a similar interest building in quality-aware least-cost Ethernet routing. Imagine feeding bandwidth, jitter, latency, packet-loss and availability criteria into a routing system that would automatically select the lowest-cost transport alternative. Instead of SLAs simply written on paper and occasionally verified, dynamic performance criteria would allow a variety of technologies and routes to serve specific applications. Imagine routing your financial, transactional, VoIP and video traffic over different, premium routes, while lower-priority email, Internet, remote storage and other applications shared a lower-cost pipe. Imagine this all being dynamically routed in real-time based on various carriers’ performance and capacity. The cost savings (and financial rewards) would certainly be non-negligible to those who could pull it off.
So how would you go about testing and verifying Ethernet link QoS on a real-time basis? Luckily the answer is already written in the standards: the Ethernet Operations, Administration and Maintenance (OAM) specs to be precise. The ITU-T Y.1731 performance monitoring standards specify one-way and round-trip delay and jitter measurements, packet-loss ratio and the ability to perform loopback-based throughput tests. Complementary connectivity fault management (CFM) OAM specs in Y.1731 and the IEEE 802.1ag provide continuity and availability data. So within OAM, all the QoS and SLA monitoring measurements are already available – just waiting to be fed into the next generation of packet-based, quality-aware least-cost routing systems.
These OAM features are gradually being integrated into Carrier Ethernet network elements, and should be predominant in the gear serving Ethernet wholesale within a few years. Today, high-performance network interface devices (NIDs) already provide full OAM functionality using dedicated-silicon, hardware-based processing – ensuring that no delay or jitter is added to the traffic being monitored. These devices allow operators to introduce OAM capabilities over “legacy equipment” that may only offer a partial feature set of OAM using low-performance, software-based solutions.
When you see the technology and a need arriving at the same time in the high-volume, cost-sensitive business that’s telecom wholesale, you know it’s only a matter of time before performance and price get squeezed to the max.
It’s just good business.
Learn all about the Ethernet OAM standards with our training poster and video at www.Accedian.com/oam.