Monotch recently completed a series of large-scale stress/performance tests for our TLEX I2V platform.
After several months of preparation, we tested almost non-stop for 2 weeks and concluded with very successful results. We were not only able to prove that TLEX I2V is efficiently horizontally scalable in very large-scale deployments, but we were also able to test several software improvements extensively and objectively establish that we achieve an even higher degree of efficiency.
Mastering data exchange
Those who have been following Monotch for some time know that we are busy connecting all kinds of roadside objects to, ultimately, the road user. In essence, this means exchanging large amounts of messages in as little time as possible. We do that, of course, reliably, safely, and with the application of the necessary governance rules.
We now dare to say that we have mastered that data exchange ‘trick’ quite well. After all, we currently deliver more than 500 million messages every day, which boasts an average of one millisecond per message. This feat is something to be proud of, but not a surprise to the Monotch team as it is fully in line with the results of the performance tests that we have carried out with every change of the platform since 2017.
TLEX to the max
But we love a challenge, and now that we support more and more types of objects and spread our wings to more and more countries, it was time for a new large-scale test session. We christened this ‘TLEX to the max’ internally.
First of all, the question was, of course, what ‘the max’ means in this case. After a healthy discussion, we decided to start from a setup of 7,500 data-providing intersections, 10,000 ‘lighter’ objects (think of road-side units, access barriers, road sensors) in combination with 5 million data-providing virtual road users and 20 ‘data-consuming’ service providers. In other words, the system delivers about 95 billion messages each 24 hours. All in all, a setup where we can support the entire Benelux as C-ITS services are rolled out on a large scale.
Co-investing in the testing phase
This was also when the financial people of Monotch began to worry. Because in addition to a test environment that makes it possible to process the planned huge amounts of data, we also needed an environment to generate these messages. After all, we can only test objectively if we generate this data outside the platform. Fortunately, our various partners were willing to co-invest in ‘TLEX to the max’, and we were able to get a green light here, albeit with a (fortunately substantial) maximum computing capacity. This somewhat changed the purpose of the tests, namely to “how much load can we process with the available test capacity?”
In addition to determining the required computing capacity for large-scale deployments, we mainly wanted to obtain certainty about the result of several improvements that we have developed for the core of TLEX I2V. That is why we composed various test scenarios in the preparation, consisting of varying loads, varying settings, and different TLEX versions.
The execution of the tests was certainly not routine. While increasing the load further, various bottlenecks were gradually confronted, particularly in the additional components such as governance and monitoring. These are valuable items that, during the test period, directly led to several reinforcements of the platform. In addition, we were able to determine that some software changes turned out to have a substantial positive impact on the efficiency of the upscaling while retaining the ‘near zero latency’ character of TLEX I2V.
After we had carried out all planned tests, we continued to increase the numbers for a while: in the end, we connected nearly 11,000 fictitious intersections with 10 million road users. Average latency? Only 5 milli-seconds ……
Bring on that C-ITS scale-up; TLEX I2V is ready!