How to interpret information from TechEmpower Framework Benchmarks?

Question

Performance evaluation

Greetings!

When studying the performance of Frameworks on the TechEmpower Framework Benchmarks site, I noticed a significant lag behind the asp.net core of popular frameworks for other languages.

For example, Weighted score:
- c# asp.net core: 6,462 (60.7%).
- java spring: 1,802 (16.9%).
- go gin: 1,667 (15.7%).
- php symfony: 1.068 (10.0%).
- php laravel: 326 (3.1%).
- js express: 826 (7.8%).
- python flask: 468 (4.4%).
- python django: 280 (2.6%).

Is asp.net core 3.59 times (Weighted score) faster than spring, and 19.58 times faster than laravel?

Answer 1

How to interpret information from TechEmpower Framework Benchmarks?

No, it's almost useless information. Look at the report of the author of fasthttp (today's fastest http framework for golang), how they contrive to score more points on this piscometer. In fact, these scores have little correlation with how your application will actually work on the X framework. And even more so, in no case should you think that "X is N times faster than Y".
Yes, on average on CPU tasksGo framework code will run faster than python framework code. But besides CPU tasks, there are also IO tasks, which are much more numerous on the web (I would say that IO tasks take up to 90% of the code running time). In addition to the framework code, there is the library code and your code that performs business tasks. And this, perhaps, is the only useful thing that this piscometer can say.

Answer 2

These are tests not of some applications, but of a number of typical actions implemented by frameworks. The contribution of actions evaluated by these metrics in real applications is not so great. Those. fetching data, and other similar work, for example, will take much more time in any request, and will be approximately the same for all frameworks, while implementing the same functionality. And this difference, it would seem at times, will turn into a value close to the measurement error.
In a real application, everything will depend much more not on the framework as such, but on the qualifications of the developers, in the end.
In addition, there is a rather complicated methodology for evaluating a pack of tests and different weights for each of them. All this can be taken into account, but certainly you should not attach too much importance to this information, and even more so, draw such straightforward conclusions.