Improve network performance by reducing the time and unpredictability required for RFC 2544 testing

Traditional telecom wires for simple phone systems are a relic of a bygone era as voice, data, and video have become mainstream and more cost-effective. Communications service providers installing Ethernet-based IP services to improve profit margins have increased service levels and demanded more advanced technology. This requires non-traditional network troubleshooting capabilities and new ways to define quality of service (QoS). In some cases, verifying quality of service is even mandated by law. In other cases, forward-thinking companies are realizing that customer satisfaction is the key to future success in their competitive industry. They need to be able to configure their networks for optimal performance, test their networks when end users encounter problems, and provide customers with full confidence in the service levels they promise. However, they also need to consider the time technicians spend testing and troubleshooting.

Service qualification is increasingly documenting network performance in the form of RFC 2544-based reports, which are determined by three key criteria: throughput, latency, and frame loss. The challenge is that RFC 2544 testing can be time-consuming, as the tests must be manually configured and take considerable time, and in some cases, the accuracy may be beyond what is needed. Therefore, communications service providers (CSPs) should ensure that they perform tests within their expectations and control, so that they can streamline the testing process to save time, but also obtain data that is accurate enough to verify performance.

When choosing a tool for network performance testing, CSPs should also carefully consider the tests that the tool performs, as RFC

2544 is not optimized for benchmarking networks and live WAN links: RFC 2544 strives to standardize single-device testing in a lab environment.

Not all RFC 2544 test equipment is equally capable of providing the critical balance between the time required to perform QoS testing and the accuracy required. Control, predictability, and automation are key capabilities required to achieve this balance. Being able to control test configurations to make testing more predictable is critical if you want to reduce the time spent testing while maintaining accuracy. In addition, the ability to build test suites can automatically test a WAN link or multiple WAN links from the same source location under multiple conditions, further simplifying the process and saving costs.

Tests based on RFC 2544

IETF RFC 2544 "discusses and defines a set of tests that can be used to characterize the performance characteristics of network interconnect devices." Rather than the ineffective analogy of "apples and oranges," RFC 2544 provides a standard way to directly compare devices. As CSPs move from traditional telecom to Ethernet and beyond, RFC 2544 as a test methodology will increasingly replace earlier QoS test methods. While RFC 2544 is a set of guidelines, not a standard, service activation testing is often based on it. RFC 2544 testing is effective for testing QoS because some customers - especially enterprise customers - require that service quality agreements (SLAs) commitments that only guarantee bandwidth include network latency and frame loss. CSPs must be able to document network performance for all of these performance attributes. RFC 2544 testing enables this documentation.

RFC 2544 Testing for Wide Area Networks

Because RFC 2544 is optimized for testing a single routing device in a lab environment, these tests are not necessarily applicable to providers installing or troubleshooting Ethernet-based IP on live WAN links. In fact, of the six tests specified in RFC 2544, only three are really necessary for analyzing performance characteristics in an Internet environment: throughput, latency, and frame loss. "Reset" and "recovery" tests are not applicable to the Internet because one cannot turn the Internet off and on. Back-to-back frame tests are used to verify that the network can support bursts of a specific duration. But this same information can be obtained from throughput and frame loss tests.

Extrapolate so that back-to-back tests do not have to be run. For example, varying the duration of a throughput test allows the user to extrapolate the measurements that would be obtained from back-to-back tests. The measurements would be the same if it were not a fixed duration. Also, WAN customers expect throughput to be present 100% of the time, not just in bursts. The following describes three RFC 2544 tests that are relevant to testing live WAN links, and potential issues with some test equipment:

Throughput is the maximum rate at which data can be transferred from source to destination without losing any frames. However, rate accuracy may be emphasized in testing, resulting in long and unpredictable tests for only a negligible difference in test accuracy. Latency is the total time it takes for a frame to travel from source to destination. The time of the relevant latency test is when the link is under load.

The network needs to be able to provide the specified latency at the specified utilization rate. RFC 2544 specifies that you must perform this test at full throughput. Frame loss rate tests are used to determine the percentage of frames that leave the source but never reach the destination. RFC 2544 recommends starting the frame loss rate test at 100% and then dropping to the next rate to test for lost frames. A successful test is determined by two consecutive tests without frame loss. Some frame loss rate tests are meaningless because they are not designed to terminate the test after finding this frame loss rate.

Table 1: RFC 2544 tests and IP-based network applications

Reduce the time spent on RFC 2544 testing

Choosing the right test equipment can reduce the time spent on RFC 2544 testing. For maximum usability, it is recommended that you use a portable, handheld tool rather than software on a PC. A laptop, while portable, is not a practical troubleshooting tool for field work. It is not designed as a dedicated test device. As such, it is not rugged enough, and the hardware on a standard laptop cannot transmit at the high data rates required to perform throughput testing. Ultimately, the ideal test equipment for evaluating and verifying network performance will reduce the time technicians spend on testing and increase control over test parameters. The test equipment should allow technicians to configure and/or control the testing to make testing more predictable and applicable to WAN links. It should support automation capabilities for testing efficiency, allowing technicians or engineers to test selected WAN links under different conditions to compare performance characteristics, or to test multiple WAN links from the same source location in a batch of test suites. In short, the equipment should be dedicated to balancing the practical issues of time and cost with the need to measure QoS by providing predictability, control, and even automation capabilities.

Predictability: Some tools are configured to favor accuracy over predictability, resulting in wide variations in the time it takes to perform a test. When technicians are trying to quantify the performance of a network or link, and you don’t know how long this will take, other tasks fall by the wayside, schedules go haywire, and customers are left in the dark about when their problems will be detected, let alone when they will be able to recover.

Said it was fixed.

Control: Control means being able to configure tools for more predictable test durations. If engineers configure tests that require shorter durations, accuracy may be reduced. However, much of this accuracy is negligible and has little relevance to real-world applications. Engineers should be able to define test parameters for new and existing networks and determine the impact of changing service levels. When configuring a new network, engineers must be able to compare configurations for optimal performance, but within a reasonable time frame. When troubleshooting an existing network, monitoring network performance and proactively identifying small issues before they escalate is much easier to do with test equipment, and you can see how many factors affect performance. However, again, this testing needs to be performed accurately and at the lowest possible cost, which means it should be within a reasonable duration. [page]

Automation: The ability to perform batch testing means that a technician can set up the tests to be run, then walk away and come back later to analyze the data. While all the parameters can be defined, it can be difficult to determine the impact they have on the network. Getting measurements means being able to optimize network performance. But running tests manually, rather than automating them, can be hampered by the time involved in testing. By setting up test suites where multiple tests are automatically run in sequence at the click of a single button, testing can be more efficient, saving time and cost, making testing easier to perform, and even giving engineers time to return to their office to define a suite and then distribute it to technicians in the field.

Introducing ES Network to improve RFC 2544 testing

ES Network Assistant II is a handheld network troubleshooting tool for real-time WAN link RFC 2544 performance testing at up to Gigabit copper or fiber cable rates. Through a proprietary adaptive algorithm, ES Network Assistant can complete IETF RFC 2544 performance testing using only three necessary tests. Through predictability, control and automation, the process is simplified to save time while still providing accurate data. ES Network Assistant is very useful for configuring new networks and troubleshooting existing networks.

RFC 2544 testing with ES Network

● Throughput: In situations where speed/accuracy is not a requirement, the technician can control the time required for testing while still providing accurate results. The required accuracy is relative. ES Network is a tool that allows the user to provide relevant accuracy within an acceptable or useful test time.

● Latency: ES Network Assistant allows the user to control the duration, repetitions, and rate of the test. ES Network Assistant allows you to change these priority values ​​and quantify them to see any impact on latency. Users can see the impact these configurable parameters have on network performance.

● Frame loss rate: Engineers select the fastest speed they want to support and start the test immediately. If any frames are lost, you can reduce the rate and run the test again. RFC2544 requires a 10% step size. However, smaller step sizes are preferred and RFC 2544 recommends a better solution. ES Network Assistant supports 1% increments. RFC 2544 defines the termination of the test as a failure threshold, not a zero threshold, meaning that the test will continue to run even if the frame loss rate has been determined. When testing WAN links, you will encounter random frame losses that are independent of the rate. ES Network Assistant can terminate the test more quickly. This is an optimization for real-world situations because it does not implement a zero threshold.

● For all three tests, ES Network Assistant supports 802.1Q VLAN tagging and provides parameter configuration for Type of Service (ToS). Users can define IP precedence or Differentiated Services Code Point (DSCP). Users can specify the same value for all tests, or change settings by test or device to test the impact on performance.

ES Network Predictability

Predictability allows service providers to provide more efficient service. By implementing control over configuration parameters, ES Networks makes test durations more predictable. A technician who has been out in the field for 15 hours is not going to want to bring a data set back to the office for troubleshooting. If CSPs can predict how long a test will take, they can schedule appointments more efficiently, providing better customer service. Predictability also makes running “what if” scenarios a practical way to test for optimal performance.

ES Network Control

ES Network Assistant gives you a level of control so that you can adjust the time it takes to complete a test by changing the configuration of the test. ES Network Assistant also gives you control over the network environment: you can configure the test to align with multiple environments related to service levels. For each test, you can use the default frame and duration defaults, or customize the configuration to observe how frame and speed differences affect network performance. You can also control the time it takes to complete a test by adjusting the test accuracy or adjusting the accuracy and failure thresholds.

The effects of changing these parameters are shown in Table 2.

Table 2: Effects of changing parameters on test time, accuracy and effect

ES Network Automation Function

Perhaps the most important feature of ES Network Communication Generation 2 is its ability to automate tests. This allows technicians to build test suites that allow them to test a WAN link under different conditions or test multiple WAN links from the same source location in one automated operation.

With EtherScope, technicians can set up a test, quickly launch, and collect a large amount of data on multiple links around the world, or data from multiple different tests on a single link. This capability greatly increases the tests that can be performed, so it can be used to set up and optimize new networks, as well as troubleshoot existing networks. Without this automation, technicians would have to set up a test, run the test, wait for it to complete, and then manually go in and set up the next test, intervening at each test point stage.

in conclusion

As CSPs install faster, more mission-critical networks and require better QoS, efficient testing will become increasingly important. Whether an enterprise is striving to prove that it meets SLAs or using performance evidence as a competitive advantage, they need the ability to configure, test, and troubleshoot their networks for optimal performance. However, given the characteristics of the 21st century economy, they must do this within time and budget constraints while still producing accurate results. Using RFC 2544 test equipment, such as ES Network Pass II, can help CSPs improve and simplify their testing processes, save time and money to conduct more tests, and gain a deeper understanding of how to configure the network to achieve optimal performance while ensuring accuracy.