If the load remains high, chances are you'll have more tasks than you're able to handle. This can cause cache misses and swapping if the requests will no longer fit into memory, as well as dropping requests and other nasty things. If you expect a heavy load, applying back pressure might be a great way to deal with it.
In essence, back pressure means that instead of putting more pressure on our service with each incoming request, we push back into the caller so it needs to handle the situation. There are a few different ways to do so.
For instance, we can block our thread that receives network packets. The caller will then see that it is unable to push the request to our service – instead, we push the pressure up the stream.
Another way is to recognize greater load and simply return an error code, for example, 503. You can model your architecture so that this is done for you by another service. One such service is the Envoy Proxy (https://envoyproxy...
