Using hardware timestamping and TWAMP for accurate, standardized network measurements
Requirements on the degree of accuracy in latency measurements are changing. As telecommunications standards continue to evolve with the rising demands of today’s consumers, so too do the latency and accuracy requirements of each consecutive generation of network technologies.
With 4G, a latency of 10–15 ms and round-trip time accuracies were standard, and acceptable. As 5G standards enter the fray, operators are finding that 10–15 ms latencies no longer satisfy the ravenous appetites of their data-hungry customer base. With 5G and IoT quickly coming down the pipe, 1 ms latencies and accurate measurements around that delay will become necessary. Use cases such as autonomous cars, enhanced mobile broadband, and critical control of robotics and remote devices, to name a few, are driving these more stringent requirements1 (but we don’t need to tell you about these use cases in detail as every single one of your telco newsletters has been referring to these for the last year). To achieve this degree of accuracy in latency measurements, hardware timestamping becomes a necessity.
Hardware Timestamping, Explored
In shared general-purpose computing systems, obtaining exact timestamps for accurate network measurements is challenging. A general-purpose operating system kernel (e.g., Linux) typically operates in a “best effort” manner with “good” average performance as the main goal. In some cases, latencies can be high and there are normally no guaranteed upper bounds. When virtualization and resource-sharing among multiple virtual machines are introduced, the problem is only exacerbated.
Modern NICs, for example those generally available in COTS x86 servers, are capable of hardware timestamping. Using hardware timestamping for both send and receive, a measurement application can improve latency measurement accuracy by 8 μs or more. Delay variation measurements can likewise be radically improved as kernel to user-space latency can vary greatly, especially under high load. Hardware timestamping also allows decoupling of the measurement application from the kernel scheduler as delay variation essentially becomes independent of process scheduling. This enables software-based network measurement equipment to increase accuracy and precision by a factor of 1,000.
Using TWAMP to Measure Round-trip Network Performance
OWAMP, which is specified in RFC 4656, provides a common protocol for measuring one-way metrics between network devices, and it can be used bi-directionally to measure one-way metrics in both directions between two network elements2. However, it does not accommodate round-trip or two-way measurements3. Enter TWAMP.
Due to its relative simplicity and reliability, TWAMP is the most popular option for reflection-based technology used to measure round-trip network performance. TWAMP has also be outlined in the IETF standard RFC 5357, driving further trust and uptake by the industry.
What a Modern TWAMP Tool Can Do for You
Employing TWAMP as a reflection-based technology brings several benefits to you as a communications service provider or network operator. In addition to the increased accuracy that hardware timestamping now provides to those that utilize TWAMP, reflector-based measurements allow you to extend your testing capabilities significantly using your existing infrastructure. A large majority of networking devices and equipment available on the market today contain embedded reflectors compatible with TWAMP and the associated protocols.
Reflector-based measurements are also ideal for the quick and easy implementation of large-scale tests, again making use of existing network equipment. Imagine, for example, a mobile backhaul scenario where there are thousands to hundreds of thousands of endpoints that require monitoring for throughput, packet loss and delay. Measuring from a test agent placed in a central location, this monitoring can be achieved without the need to place receiving test agents on all of these locations, saving time, resources and substantial costs.
Further benefits can be found in the ability to remotely troubleshoot network problems without having to send technicians into the field. Reflection technology and segmentation of the network allows you to speed time to issue isolation and proactively find problems before customers notice any service quality degradations. Dispatch technicians to fix problems as needed, not to find them.
A final item to note is programmability and the capability to automate this testing and monitoring using APIs. Using a network orchestrator, such as Cisco NSO, or other OSS system, activation tests, ongoing monitoring, and even troubleshooting sequence can be automated, furthering your costs savings, decreasing time to deployment of new assured services, and eliminating the human error factor.
To learn more about our research around hardware timestamping and TWAMP, download our technical whitepaper. This white paper goes into deep technical detail around both the practice of and the benefits derived from hardware timestamping in Linux-based software for network service quality measurements. Data from network interface hardware datasheets is compared with actual measurement results. The research was carried out from a Linux measurement application using hardware timestamping in conjunction with the reflector-based TWAMP. The white paper also provides a short overview of how hardware timestamping can be used in virtual machines (VMs).
Interested in learning more about hardware timestamping, TWAMP and how Netrounds is using these technologies to help our customers take advantage of automated active testing and monitoring? Contact us at firstname.lastname@example.org to schedule a one-on-one demo.
Download our technical white paper here.
1Ericsson Mobility Report, June 2017 https://www.ericsson.com/assets/local/mobilityreport/documents/2017/ericsson-mobility-report-june-2017-rina.pdf