The software industry is largely dependent on integration and traditional alert tests to prevent issues and discovery. This is behind the internal standard set by major technology companies, which use advanced technologies to detect problems long before the impact on a large group of customers.
Discovering a problem earlier means that you reduce or even get rid of the customer’s effect. In this article, you will learn about some of the most effective issues to detect issues and when they publish veteran technology companies.
Synthetic monitoring
If you only have a few customers, it is difficult to measure if the product works properly. Keep in mind that you only receive a few transactions per hour on your website. If the transactions decrease to scratch, you do not know whether the product is broken or if customers are inactive at that time.
Artificial monitoring is the solution to this problem. Synthetic monitoring is the use of a simulation customer to investigate the system and measure the results. The probe should implement all steps just like the real customer and should work at a fixed frequency. You should alert your monitoring system if the artificial failure violates a specific threshold.
It can be said that artificial monitoring is the mechanism of discovering issues and prevention that you can get-more important than comprehensive tests or tests. This is because the best way to check the final experience of real customers is to create industrial customers.
Canary test
Canary is more sensitive to the first carbon monoxide, so it was used in charcoal mines to determine whether the gas is present. If the Canary gets sick or dies, the miners evacuate the mine. For internet, this is parallel to exposing a new feature or changing to a small group of users (“Canary”), and monitoring whether these users are negatively affected by the change.
The Canary test requires you to measure the experience of Canary users of deviations in cumin and availability. If the scales are decomposing to the canary, you should thwart the change and install it automatically. If no deterioration is observed, the change should be automatically to be gradually brought up to a group of clients.
The Canary test should be one of the start -up pillars, but it may not be particularly useful when the customer traffic is very low. However, most modern operating tools support Canary test, so it is still good to prepare it early.
Shadow test
Imagine that you are an application to deliver food and release a new algorithm to choose a driver who improves delivery times. These are commercial important functions, so you want to be very careful.
In the shadow test, the new algorithm is operated along with the original group of a sub -group of traffic, but the original algorithm will continue to choose the delivery. You can record the results of both algorithms and their comparison. If the two algorithms do not agree to the majority of the time, you must likely investigate whether the new algorithm chooses the appropriate delivery drivers.
The shadow test is a great tool to use when your product has a lot of exchange that is validated through traditional tests is impossible. In our delivery example, not all nuances in the real world can be repeated, and thus comes the shadow test for rescue.
Note that the shadow test does not prove that the new algorithm has better delivery times. Once you are validated that the new algorithm gives “reasonable” outputs, you should do a A/B test to confirm that it actually improves delivery times.
Automated download test
The worst moment to know that you cannot handle the scale … is when you need to deal with the scale. The usual practice of high -motion products is an automatic download test that is run before any change goes to production.
This includes the formation of the download test environment that can simulate production and create an artificial traffic that pushes the system to its borders. Then you need to determine the success parameters for the download test. There is an initial way to do this, which is the baseline of the resulting metrics (cumin, availability) against the oldest and successful download test. This will also help you get rid of the changes that cause measurable deterioration in the performance of the system.
conclusion
The software industry has long relied on the “test pyramid”, which contains unit tests at the bottom and integration or tests from one side to the party at the top of the pyramid. The test pyramid is desperately old in a world where most services are built from fine services, and most programs are accessed online. In the modern context, the practical benefit of technologies such as synthetic monitoring or canary test is much larger than integration tests.
In this article, we have explored some of the most effective technologies to capture issues early, although they are definitely not a comprehensive list. Depending on the nature of your system, you may also want to assess the discovery of anomalies, failure injection or height test. More importantly, benefit from all power outages to investigate the detection mechanisms that lack it, and fix the gaps that you find.
