Carrier Ethernet Testing
Like in many other areas, Ethernet is poised to dominate WAN access. Ethernet offers many advantages over traditional TDM services. Most obvious is Ethernet’s Low Cost. With so many vendors and so much competition, prices are aggressive and dropping. For customers, Ethernet can often be very inexpensive to connect since they likely have unused ports on existing equipment. Bandwidth Granularity allows carriers to offer the exact bandwidth a customer desires and scale up easily. For example, a carrier can install a 1GBps link and limit the bandwidth to 200MBps. If the customer outgrows this limitation, a simple reconfiguration can change to limit to the required value. Traditional TDM circuits required purchasing bandwidth in larger increments, and required completely different hardware to move, for example, from T1 to T3 to OC-3. That brings up the fact that Ethernet is scalable to high speeds – up to 1GBps and higher. Ethernet also offers greater simplicity. The Ethernet frame that leaves the customer premises can be routed quickly and efficiently through an all-ethernet carrier network.
MarketThere are a variety of applications where you will find carrier Ethernet. Retail sales refers to cases where a service provider sells Ethernet services to an end customer, typically a business of some sort. Traditional telecommunications providers are strong in this space, but they are now being challenged by new, specialized providers, and MSO’s. Multinational customers usually want a single service provider to solve all of their communications needs. However, very few service providers have a presence in every market in the world, so if they want to offer end-to-end service, they need to contract for Ethernet services from local providers. This sort of service-provider-to-service-provider purchase is called wholesale. Ethernet testing and troubleshooting for these applications can be particularly challenging as the carrier doesn’t have insight into the entire network.
Backhaul refers to non-public networks used to interconnect other communications technologies. The largest market for backhaul is in mobile telephony. Many of the connections between cellular stations and central offices are carried over Ethernet, and that market is expected to post significant growth. Other examples of backhaul include WiFi and WiMAX hotspots and satellite communications. There are also users of Carrier Ethernet who are not service providers. Some customers lease fiber connections that are nothing more than a physical connection between sites and install their own equipment at each end, which is often Ethernet. This is known as buying “dark fiber”. And many large, geographically concentrated organizations such as military bases, city governments, and utilities install and manage their own Ethernet networks. Some of them even sell Ethernet services to other organizations.
Turn Up TestingWhen a carrier installs an Ethernet circuit, there are a number of ways to make sure that it is functioning properly. The simplest ethernet test is to simply look at the lights on the equipment and see if they light up. A slightly better ethernet test is to connect a laptop and PING a router or perhaps surf to a web page. These show that the link is working, but they don’t give any indication of the actual performance of the link. Customers who are paying high process for advanced services may want some better indication that they are getting what they are paying for. And carriers need to verify the circuit is meeting any Service Level Agreements they may have with the customer.
BERT is an abbreviation for Bit Error Rate Testing, which is an old telecom testing method where frames are sent across the link at the maximum possible rate and errors are counted. This shows the customer they are getting the bandwidth they are paying for. The most complete ethernet test is RFC 2544 - a specification developed by the Internet Engineering Task Force, the same guys who invented the web, email and other modern conveniences. It defines a way to measure all the critical performance characteristics of a circuit. The first is measurement is throughput, basically the maximum frame transmission rate, which, for ethernet can range from 2 megabits per second to one gigabit per second and beyond. Second is loss – how many frames the network loses between one end and the other. Burst testing is third and it characterizes the circuit’s ability to handle frame rates above the specified maximum. For example, a service provider may provide a 20 megabit circuit that has the ability to pass 50 megabits per second for short periods of time (less than a second). That way, the customer doesn’t have to pay extra for bandwidth that they need only rarely. Ever had a phone call where it took a long time for the other person’s voice to get to you? That’s the fourth parameter – latency or delay. And just like latency can make it hard for you to talk, it makes it hard for computers to talk.
Finally, jitter is not part of 2544, but it is often tested at the same time. Jitter in the variation in latency. On modern networks, the time it takes for frames to get from one end to the other can vary. Jitter is the measurement of that variance. Loss, jitter, and latency are not especially critical for most applications. But in applications where the data has to appear in real time, such as voice over IP or streaming video, there is not enough time to retransmit or reassemble lost or delayed frames. So having acceptable levels of loss, jitter, and latency are important. Service providers may not use all these tests, but if they do, it provides the most complete characterization of a circuit.
Monitoring and TroubleshootingTurn up testing is a special case in that the link is not operational when the tests are performed. Once the link has been turned over to the customer, this sort of test can only be done on rare occasions when the link is taken out of service. There are ways to test a link that is in service as well. The RFC 2544 tests noted above can be used to characterize a circuit, but devices on each end must be connected in a manner that allows the customer traffic to flow unimpeded. Further, the tests that are done should not place undue burden on the link. For example, attempting to load the link with the maximum traffic level will likely result in unacceptably poor performance for the customer’s data. To prevent this, traffic levels need to be kept low. Another approach is to place a very low priority on the test frames so that the network will discard them first. These approaches will allow measurement of all RFC 2544 parameters with the exception of throughput and possibly loss, which, by their nature, require loading (or overloading) the link.
Sometimes troubleshooting requires monitoring of the actual customer traffic on the link. There are a number of approaches for doing so. The first is to use a two-port tester, which can be connected “in line” with the traffic. This, however, requires disconnecting the link which may not always be possible. The second is to use a tap which can be installed on the link and left for connecting whenever the link needs to be monitored. This is an additional cost, but tap prices are usually a tiny fraction of the cost of a tester. A third way is to let the network equipment tell you what is happening. Testers with SNMP capability can talk to the switches and routers in the network and show the operator what they are seeing. While this does not offer the depth of analysis of an in-line or tap connection, it is much simpler as there is no need to break the link, nothing to install, and it can be done from anywhere in the network.