Connections +

New Age Networks

With prices dropping and test equipment available to help install and maintain them, the future of native Ethernet over fiber is sure to grow. Agilent's Peter Schweiger explains how....

July 1, 2006  

Print this page

With prices dropping and test equipment available to help install and maintain them, the future of native Ethernet over fiber is sure to grow. Agilent’s Peter Schweiger explains how.

In the race to reduce their cost per bit transported, telecommunication and broadband service providers are embracing two new technologies: Gigabit Ethernet (GigE) over Dense Wavelength Division Multiplexing (DWDM) transmission, which places multiple 1Gbps Ethernet links over a single fiber using multiple wavelengths and allows more data over fewer fibers at much lower cost then SONET links.

Since the wavelengths are within the passband of optical amplifiers, this combination can work between nodes, across town or even cross-country.

Let’s look at why and especially what needs to be tested in a business model that is growing as we speak.

Fiber is everywhere. IP is everywhere. In a race to provide triple play services to consumers and business Ethernet connectivity, service providers are accelerating their deployments of Ethernet connectivity and Ethernet transport throughout their networks from Access, through metro to the core.

Three developments are enabling this transformation:

* Gigabit Ethernet (GigE) links have come down dramatically in cost, and continue to fall to prices below $1,000. They offer easy, familiar connectivity and high bandwidth at a fraction of the cost of SONET transmission systems.

* Fiber networks can handle high aggregate data rates effortlessly and are being built closer to customers with many receiving a fiber based GigE right to their premises.

* The ability to combine GiGE links at multiple wavelengths on a single pair of fibers to satisfy bandwidth needs in the metro and core networks quickly even with only two fibers available.

IP knowledge lacking

Some testing techniques are understood, but IP lacks the ‘5-9’s (99.999%) maturity and experience that voice transport has. On top of this as IP connectivity moves to access networks, technicians with little IP knowledge need to provision and troubleshoot networks.

Since fiber testing is well understood and doesn’t fundamentally change as it pushes closer to the network edge, we will focus on explaining testing of the other two enabling technologies: Ethernet and multiple wavelength systems.

Basic Ethernet testing: From 10 to 1000Mbps: Basic Ethernet connections requires two major types of tests to guarantee Service Level agreements with customers (SLA’s) or ensure a GigE is reliable enough to transport traffic node to node.

All Ethernet test sets send and receive frames to ensure that end to end connectivity is established. This can be done with a MAC level or IP loopback between test sets. Sometimes only a simple PING test is done. In 1999 the Internet Engineering Task Force (IETF) created a document called Request for Comment 2544 (RFC-2544) with suggestions for testing point-to-point Ethernet links.

Now service providers could benchmark performance consistently for future comparison and provide a way to measure QoS to maintain an SLA with customers. Also, customers could objectively compare service from different providers easily.

Below is a description of the four key tests recommend during field-testing.

Bandwidth: Measured by generating traffic at the maximum rate, called full line rate, across all supported frame sizes to determine the maximum throughput in % of link speed or in Frames/Second.

Latency: Also known as network delay, it is important to VoIP applications and is measured by loading the network and measuring round trip packet transfer time between test sets. Round trip values of greater than 200ms will affect VoIP quality and greater than 300ms round trip are noticeable by users.

Frame Loss: Or rate is similarly done by loading the Ethernet link from 0 to 100% and expressing any frames that didn’t arrive as a percentage of the total.

The ‘Burstability’ of the network is found by counting the maximum number of frames transmitted immediately after each other at the maximum allowable speed that can be transmitted without errors.

Complete tests can take anywhere from 35 minutes for a useful test to over five hours for a complete stress test with multiple iterations and longer durations recommended in the RFC-2544 document. Results can be expressed in a table or graphic form and are increasing stored and shared with customers in a report.

Two important things to note about these tests: First, they are designed to measure point-to-performance and not end to end connectivity through multiple routers and security devices such as firewalls and Network Address Translation devices (NATs).

Secondly, all the tests generate full load conditions on the link and care should be taken to perform these tests when the link is not yet in service or carrying critical traffic.

When testing point-to-point Ethernet performance, choose test equipment that can perform the RFC-2544 series of tests, and execute them easily.

For most GigE links augmenting SONET transmission as a transport connection across service provider networks, the basic Ethernet performance testing recommended in RFC-2544 is sufficient.

But for Ethernet links meshing with end user enterprise networks, going beyond Quality of Service to measure Quality of Experience has proven to be an opportunity to increase customer satisfaction, loyalty and reduce the number of service calls when issues arrive.

Measuring QoS involves two steps. Identifying key networks resources that customer’s access and measuring those resources in a repeatable way that simulates actual use. A bonus is the ability to see a breakdown of the different steps to locate the cause of problems.

Why go to this extra trouble when generally a service provider’s responsibility ends showing link performance? An example from a service provider’s perspective might help explain the rationale: An important but new customer calls to say that their Web server access connected by a link from a service provider is extremely slow. They suspect the link. The provider quickly rolls a truck dispatching a technician who goes and checks the link even interrupting service briefly to test it fully. The link shows no faults. The customer escalates the issue raising tempers, the number of people involved and technician’s mileage.

With new test tools the provider’s technician can connect to the customer’s network, quickly access the (web server and rate the performance of each step involved in using that resource.

The results can be rated poor to excellent (or 1 to 5) and the individual steps can be broken down to determine where the problem lies. In this example, the network connection worked perfectly, but the response time of the web server itself was the source of dissatisfaction.

Instead of escalating frustration, the technician becomes a hero, the customer satisfaction is restored and expectations are even exceeded. In short, the customer is delighted.

Such a use of network node Automatic discovery can locate rogue stations, such as unauthorized wireless access points. A check of network utilization can determine if a workstation has a defective card or virus causing it to flood the network with broadcast frames and locate it.

Becoming DWDM Aware: In order to get increase bandwidth inexpensively and achieve necessary bandwidth from only two fibers, Wavelength Division Multiplexing (WDM) has been employed.

Transmitters (for example GigE links) operating at predetermined wavelengths defined by the International Telecommunication Union (ITU) are combined by a wavelength division multiplexer (Mux) coupled to the fiber available and connected to a wavelength division de-multiplexer (DeMux) at the far end before being connected to receiver ports. The ITU specify bands from 1270nm to 1675nm but the advantage of using the 1530nm to 1565nm ‘C’ band is that those channels have the unique
ability to be amplified, regardless of it data rate, by optical amplifiers.

The exclusive test tool of choice for all the required measurements is an Optical Spectrum Analyzer or OSA. Most feature one button operation populating a table with all the required measurements. Some can even operated from a battery and are light enough to wear when using in a confined space.

Tests are preformed at multiple points along the link to check levels and isolate problems, like a failed transmitter or bad amplifier. A key note is that you cannot test what you cannot connect to, so wise network planners add low loss 1:20 couplers to provide monitoring points to hook up OSA’s without disrupting traffic.

With prices dropping and test equipment available to help install and maintain them, the future of native Ethernet over fiber is sure to grow.

The good news is that neither Ethernet nor any optical tests change as speeds increase. Although still less than 10% of the market, 10 Gigabit Ethernet is poised to repeat the value proposition offered by the current GigE explosion.

Rest assured test equipment vendors are working to bring down the price and size of test equipment to match customer’s budgets and expectations as they move to embrace IP over fiber.

Peter Schweiger is the business development engineer for Agilent Technologies Inc.’s Photonics Measurement division.