Npgsql.FSharp VS ply

When working with F#, my database layer code has always seemed ugly. It feels like functional languages don't pair well with storage. I've used F# with Dapper, EF, and Npgsql.FSharp. Although Npgsql.FSharp works well, I would still like to use something like EF since it handles inserts and updates so well.

Not a big fan of ORMs in general, I've been using https://github.com/Zaid-Ajaj/Npgsql.FSharp for postgres.

On Postgres, there's a decent low ceremony option by Zaid Ajaj: https://github.com/Zaid-Ajaj/Npgsql.FSharp

For database, I use PostgreSQL with https://github.com/Zaid-Ajaj/Npgsql.FSharp and I like it very much. Writing the services, I do hide database operations behind interfaces like

Npgsql.FSharp

I work with mass-scale concurrency using green threads (~50000 server tasks a second, as many threads at a time as possible, ~50ms latency differences matter). How is the performance relative to something like Ply? I am particularly concerned with GC pauses.

Ply is a great library for interacting with Tasks: https://github.com/crowded/ply

DISCLAIMER: I'll be using Ply in most of the samples due to it's efficient task CE and the easy interoperability between C# tasks as well as F#'s async. Http operations are by nature async operations, that means using async/await in C#/JavaScript or using async {} in F# which can be a little annoying when you need to append |> Async.AwaitTask to every function/method that returns a Task.