Re: Extensibility of the PostgreSQL wire protocol

On 19/2/21 14:48, Heikki Linnakangas wrote:
> [...]
>
> I can see value in supporting different protocols. I don't like the
> approach discussed in this thread, however.
>
> For example, there has been discussion elsewhere about integrating
> connection pooling into the server itself. For that, you want to have
> a custom process that listens for incoming connections, and launches
> backends independently of the incoming connections. These hooks would
> not help with that.
>
> Similarly, if you want to integrate a web server into the database
> server, you probably also want some kind of connection pooling. A
> one-to-one relationship between HTTP connections and backend processes
> doesn't seem nice.

    While I'm far from an HTTP/2 expert and I may be wrong, from all I
know HTTP/2 allows to create full-duplex, multiplexed, asynchronous
channels. So multiple connections can be funneled through a single
connection. This decreases the need and aim for connection pooling. It
doesn't eliminate it completely, as you may have the channel busy if a
single tenant is sending a lot of data; and you could not have more than
one concurrent action from a single tenant. OTOH, given these HTTP/2
properties, a non-pooled HTTP/2 endpoint may provide already significant
benefits due to the multiplexing capabilities.

    So definitely we don't need to consider an HTTP endpoint would
entail a 1:1 mapping between connections and backend processes.

    Álvaro

--

Alvaro Hernandez

-----------
OnGres