one request = one h2 stream

Key Takeaways

But what does it What does it mean in HTTP In HTTP there is no such limit. There is no such pool in the client side. There is a maximum concurrent stream that can be negotiated between the server and the client. The server say, "Hey, here's the maximum number of concurrent streams. I I e if you have This is how many streams can be open at a given time, which kind of you can think of it as stream as a request in HTTP 2. In this particular case it's it's the same. Now gRPC overloads that and uses it for bidirectional stuff and way more features, but let's keep it simple, request response. We love that stuff. So, that default is I think it's a hundred, sometimes two hundred. But hundred is still a lot. So, that is why if you ever run into this problem before I mean when when HTTP got popular the moment you switch HTTP 2 is just a bit. That say, "Hey, server now supports HTTP 2." You just advertise that in your ALPN during TLS and the client Chromes of course supported, your websites all of a sudden sometimes you start seeing failures. Why? Because the client will just send request. There is no pool anymore. It's just Oh, I can send a hundred request? When I used to send six at the time, I can send a hundred? It will just send. There is no queue anymore for a connection pool. That it doesn't exist. It's one connection and then you We can send all these requests at the same time. And what happens is back end started getting overloaded by just turning that bit. Many examples on the web you can see this. And of course the solution is add more hardware to the back end, spin up more servers. That works, but without understanding the core of the problem you'll one you'll be you'll left wondering what happened. And to me that's not the way I work as an engineer. I always want to get to the bottom of things.