Why test your 40/100G network? 10 things you need to know

The exponential growth, expanding bandwidth demand and delivery requirements in global networking for video, voice and data applications are accelerating at a speed never before seen. For many, the original transport backbone architecture built on 10G wavelengths is at the threshold of capacity. Higher-speed 40G and 100G are the next-generation transport solutions that enable and define future success for service providers, network equipment manufacturers and a growing number of enterprises. Recognising the right steps of due diligence, which include testing, can ensure the delivery of all the benefits expected of this game-changing solution.

Q: On a high level, I understand the growth of demand that pushes the need for 40/100G. Can you give me a deeper insight into the market trends?
A: As telcos compete with cable providers, with IPTV and on demand use increasing, the bandwidth requirements are also exploding. HDTV uses over five times the bandwidth of SDTV (19 Mbps vs 3.5 Mbps). YouTube has experienced a 20 to 50% monthly growth rate. In wireless, video and other applications are constantly pushing the limits of the backhaul network. Social networking is responsible for spikes of usage as people increasingly upload full-resolution images from low-cost digital cameras. Storage and data centres continue to grow and data centre virtualisation represents a critical increase in inter-server traffic as processes are spread across multiple physical machines. The equities trading industry has increasing demand for massive volumes of data (billions of transactions and terabytes of information per day) where low latencies are a matter of survival and surplus bandwidth is essential.

Q: I need more details about what aspects of scalability and performance are the critical points of 40/100G testing.
A: Scalability - scale to the highest physical (ports) and/or virtual (emulated subscribers and devices) density; performance - full line rate traffic generation with the highest connection rates for application layer traffic and emulated protocols.

Q: We are planning to manually test our 100G solution before launch with our small team of experts who have studied the associated technology and methodologies. We have the right up-to-date test equipment for the job. Are we missing anything?
A: Due to the significant increase in bandwidth associated with the move to 100G, there is a corresponding increase in test results to be analysed. More of everything must be supported - more routes, more VPNs, more tunnels, more queues. Because of this it becomes more important to keep the number of test runs to a minimum without sacrificing test coverage and product quality. Hence, if manual testing has been the strategy in the past, it may not be suitable to handle the needs of the future. Automation is an essential element in reducing test time, especially with a small test team, and particularly with the greater test demands imposed by 100G. Test case automation, which uses Tcl scripts or saved configurations to quickly configure and run tests without intervention from the lab engineer, is the most common form of automation. The benefits and advantages are critical in that they free the engineer to do other critical work, foregoing the repetitive requirements of manual testing. Another advantage of automation is 24/7 testing, which maximises test time. Other opportunities for automation also exist, such as test lab automation. All automation solutions help you speed your 100G solution to market, with accurate results that help you ensure the quality of service you and your customers require.

Q: What network performance tests are recommended for 40/100G?
A: RFC 2544 testing, which measures latency, jitter, inter-arrival time, frame loss. RFC 3393, which calculates real-time jitter as required by MEF-10 and RFC 5481. In addition, sequencing, as well as PRBS bit error rates should be tested. In lower speeds this is not an issue in testing. In 40G and 100G, however, this is critical. For example, as the length of cable increases we see packet drops, even in a straight piece of cable. In this stratum of testing, bit errors happen much more frequently, and accounting for error rate speed is critical in 40/100G testing.

Q: What are quantifiable benefits of performance testing my 40/100G network before launch?
A: Testing can verify network equipment manufacturer product claims, acquiring consistent, comparable statistics to provide objective comparison of each tested vendor. In addition, not all networks will be composed of equipment from a single vendor, so testing allows you to find performance data between different vendors. You are able to establish network/device performance under load, identifying break points and allowing issues to be resolved prior to the subscriber experience.

Q: I recognise the need to test my 100G solution deployment for performance and scalability issues and the need for up-to-date equipment. I’ll just have my in-house team get up to speed on the technology and methodologies and take over with the testing guidelines I learned about here, right?
A: Best practices clearly indicate that testing before a 100G network deployment goes live is an effective practice to mitigate costly risks. However, relying on in-house engineering resources, which may lack the required up-to-date expertise and hands-on experience, can produce unreliable test results. The choice of a qualified test partner is critical, especially when deadlines are short, budgets tight and margin for error is zero. Employing a partner with a team of test experts conversant in the challenges in the field of 40/100G testing is essential. Also, since the cost of the required test equipment can be prohibitive, selecting a partner that also provides rental options offers added advantages.

Q: What specific qualifications should I look for when choosing a third-party test partner?
A: First, the partner must be a neutral party, without a motivation to influence the test results, as may be the case with a system integrator testing their own solution. Some systems integrators who do their testing in-house have a vested interest in delivering positive results. Second, testing should be a core competency of the partner, not an ad hoc solution offered on request. This means the partner is an established global name in the test and measurement industry with verifiable experience and expertise and a team of quality test engineers. The testing team should have a holistic understanding of networks, be able to articulate testing benefits and ROI vs risk, and have an established delivery process. A test partner’s qualifications are further strengthened if they have extensive expertise in lab and test automation and can provide references of successful engagements. The most advantageous test partners that provide test rental equipment are those who supply devices manufactured by the same company.

Q: We have very reliable performance testing gear in our lab which we have used with great satisfaction over the last three years. We intend to employ this platform for testing our new 100G solution. Does this testing strategy make sense?
A: Resolution of legacy test equipment (including those for 10GE testing) is 20 nanoseconds (ns) or more, which cannot deliver accurate measurements for 40/100G. Here is why: Assume one frame every 672 bit times, a 12-byte inter-frame gap, an 8-byte preamble and the smallest frame size of 64 bytes. The time between two frames is 67.2 ns for 10G, 16.8 ns for 40G and 6.72 ns for 100G. Therefore, if the traffic generator has a clock resolution of 20 ns it will send two frames at 40G and three frames at 100G with the same Tx or Rx time stamps. One must have at least 16.8 ns resolution at 40G or 6.72 ns resolution at 100G to uniquely timestamp contiguous packets. Having the right test equipment with the right nanosecond resolution is essential to test properly and get the results you need to ensure your 100G solution is ready for launch.

Q: We are planning to migrate to a 40G solution. My highly qualified team of engineers has ample experience with designing and implementing 10GE networks. If we have their stamp of approval, is it really necessary to test our newly proposed solution any further before launch?
A: First, 40G is very different from 10G technologies. It’s not simply a matter of being four times faster than 10GE, but a multilane, parallel wavelength system. Jitter and sequencing testing may not be that essential to a 10GE network but they are critical to multilane systems. Having a team of engineers conversant with 10G networks does not guarantee their expertise in 40G and testing properly with a suitably qualified team before launch addresses that issue.

Q: Our budget is approved for upgrading our testing platform for our planned 100G network. Are there other components I need to keep in mind as well?
A: Not all system clocks are created equal. Thorough research is essential to ensure the one you have is right for your testing requirements. A seemingly insignificant detail such as clock resolution can affect the ROI of your test lab investment as well as the TCO (total cost of ownership) for the equipment purchased. The time to ask the hard questions, such as can the system scale to test routing at high speeds, or stream and analyse real video, is before the test platform purchase is made.