Replies: 10 comments 6 replies
-
|
I understand where you are coming from, but are those architectural concepts really transferable, especially between different languages? It might be challenging to classify implementations consistently and uniformly. |
Beta Was this translation helpful? Give feedback.
-
|
I don't see why they wouldn't... Can you elaborate? |
Beta Was this translation helpful? Give feedback.
-
|
Well, take "threads" as an example - in one environment the term may mean entities that are scheduled by the Linux kernel, while in another it may mean what is commonly known as fibers (or green threads). A third group of runtimes may actually implement M:N threading, which is like a hybrid of the other ones. So, if you say that an implementation is multi-threaded, for example, what do you mean exactly?!
As the author of a C implementation, One way I could see it working is if the architecture is described from the viewpoint of the operating system, but I suspect that this is not the level of detail that you are looking for - for example, it is possible that some of your categories might appear to be exactly the same to Linux. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for providing more details @volyrique (and sorry for my huge delay 😅)! I thought about the following matrix of parameters to define them:
Not sure how confusing it might be... I'm just concerned with the idea of comparing, for example, Symfony+Nginx+PHP-FPM with Symfony+PPM (a multi-processed web server implemented in PHP using fibers) with Symfony+FrankenPHP (PHP embedded in a Go webserver). I'm sure, though, that all supported frameworks would possibly have different results on the benchmark depending on the available architectural categories. |
Beta Was this translation helpful? Give feedback.
-
|
Exist a long list of options, and sometimes they are blurred or even have various. For example, FrankenPHP use a different model if it's working in worker mode or not. And most options, less than 2% normal users will understand it.
That's the idea of the benchmark, compare frameworks using any architecture. Symfony
Latency Also we have the classification columns, than we can filter to show only for specific options. Type of framework, language, platform, server, ....
We only need to filter to view only the preferred options. Example to only show the frameworks using fpm/fastcgi I tried for some years that the plaintext test show in the classification, the frameworks using http1.1 pipeline. And still is not added. |
Beta Was this translation helpful? Give feedback.
-
|
What I'm trying to convey is that I believe it'd be useful to have a more generic classification for things. For example: being able to compare nodejs and python solutions with php ones that follow the same architectural design. |
Beta Was this translation helpful? Give feedback.
-
|
You can do it, with the filters !!! |
Beta Was this translation helpful? Give feedback.
-
|
I prefer say stack (architecture) |
Beta Was this translation helpful? Give feedback.
-
|
Those were just examples... the issue with the filters is that people need to know about the architectural details of the platforms in order to achieve what I'm describing. Say someone wants to compare everything of all languages that is multi-process based with native http server, or everything that uses hybrid threading... idk It's fine if that's not a problem in opinion, btw 😁 |
Beta Was this translation helpful? Give feedback.
-
|
Nodejs only have one architecture, Python have a lot, ... and PHP is new having new architectures from the classic fcgi shared nothing. |
Beta Was this translation helpful? Give feedback.
-
|
It will be not easy for me, to try the correct architecture with python. But I'll try my best. Edit: with PHP will be not easy too!!! |
Beta Was this translation helpful? Give feedback.
-
|
First, when you find the people that understand it, please say me !!! PD: thank you !! |
Beta Was this translation helpful? Give feedback.
-
|
This benchmark have more than 400 frameworks, with more than 300 platforms, .... |
Beta Was this translation helpful? Give feedback.
-
|
Nobody is interest about the technical details, they only want use less servers for serve their users.. |
Beta Was this translation helpful? Give feedback.
-
|
Looking at the architectural models allow us to comparing apples to apples and pears to pears, regardless of language or stack, which makes the benchmark more accurate in general. Let's just say it's not a concern for now, then. Thanks for your input 👍 |
Beta Was this translation helpful? Give feedback.
-
|
Only PHP with any platform !!! |
Beta Was this translation helpful? Give feedback.
-
While implementing improvements (#8550) and having discussions with some folks, I observed that we can increase the usefulness of the benchmark model by better categorising the architecture model used (instead of simply the framework).
Let's take PHP frameworks as example, the benchmark compares the mainstream model (share nothing architecture), with "native" async via event-loop, with "hidden" async with extensions, with "process/worker managers" based things.
That makes it difficult to separate "framework overhead" from "architectural constraints".
Expanding the perspective to other languages, causes this "apple vs pears" effects even more visible (we add "multi-process" vs "multi-thread" vs "multi virtual thread" to the mix).
Other discussions are pointing to the same thing but (IMHO) attacking the symptom and not the root cause.
My proposal would be to provide additional details regarding the architecture model on the implementations, which would also allow us to see the different consequences of using "framework z" against different architectural models.
Beta Was this translation helpful? Give feedback.
All reactions