How database queries are works?

[eugene-nuwber]

Hi, I'm curious about how Eloquent handles concurrent queries in Hypervel, specifically with eager loading. The documentation shows an example with multiple relationships. If I've already fetched 25 IDs and then eager load several relationships for these models, will Hypervel execute the relationship queries concurrently (in parallel) to PostgreSQL, unlike Laravel which typically does them sequentially? I'm wondering if this would provide a performance benefit in this scenario within Hypervel?

[albertcht]

Hi @eugene-nuwber ,

These queries won't be executed in parallel unless you make them run in parallel. For the usage of coroutines you can refer to coroutine in the docs.

If you don't have experience in coroutines before, I strongly recommend learning about Linux operating systems, asynchronous I/O, coroutines, network communication protocols, concurrency, and other fundamental knowledge through Mastering Swoole PHP. This will help you gain a more comprehensive understanding of how Coroutines work behind the scenes when you use them in Hypervel.

[eugene-nuwber]

I've already read this, thank you.
My first try was to use the parallel function and everything works fine. Then I found the sentence which is telling that the entire server is coroutine and this made me think that my queries can be ran in parallel. So thank you, now I understand that I should do all requests manually instead of using Eloquent features.

I'm sorry, but it this case Hypervel has no profit in terms of using Laravel Octane. Of course you provide the Coroutine ability and let us to do parallel queries if we want (Hyperf using if preferred then because it not confuses), but... why don't to do this out of the box, it looks easy to implement.

[albertcht]

1. I/O operations won't be executed concurrently in coroutines unless you make them run in parallel. The ideas of coroutine and concurrency don't conflict at all.

Then I found the sentence which is telling that the entire server is coroutine and this made me think that my queries can be ran in parallel.

2. Don't mix the concepts of QPS performance and response latency together. In Octane you can still send requests concurrently using processes, but that doesn't change the fact that each worker can only serve one request at the same time. On the contrary, if you send multiple requests during the request lifecycle sequentially in Hypervel. The total time of these I/Os will affect your response latency, but the worker can still serve other requests at the same time while waiting for I/Os.

this case Hypervel has no profit in terms of using Laravel Octane

3. The behavior of eager loading can also be found here: https://github.com/hyperf/hyperf/blob/master/src/database/src/Model/Builder.php#L693-L705

4. The control of the concurrent behavior should return back to the developers. Concurrent queries to the database can reduce the response latency, but can increase the pressure to your database as well. Sending queries concurrently in the background seems hiding the details of the control of your I/Os, which is dangerous.

Of course you provide the Coroutine ability and let us to do parallel queries if we want (Hyperf using if preferred then because it not confuses), but... why don't to do this out of the box, it looks easy to implement.