Benchmark with PostgreSQL instead Cockroach

Hi there!

Someone knows if there are any benchmark with PostgreSQL instead Cockroach?


Hi @luizvaz. What sort of performance requirements do you have? We’ve scale tested Nakama with Postgres, CockroachDB, and AWS Aurora at real-world usage levels. The server is in production use with games which are top 100 on mobile.

Is there a specific set of performance numbers you’re interested to check and on what size of hardware?

Hi @novabyte.

I was just wondering how it compare with the docs
In my case, I don’t see more than 1000 CCU.

Did you know any use case of open world?

@luizvaz It doesn’t look like you’ve read the benchmark page properly. The 1000 CCU is on a machine which runs Nakama on 1 CPU core and 3.75 GB of RAM. This performance is fantastic and likely has improved even further since the benchmarks were published.

How many CCU do you expect to handle per server?

Did you know any use case of open world?

  • How many players per region of the open world?
  • How will you divide up the world space?
  • What sort of multiplayer experience is the open world game type?

“open world” is a broad topic for multiplayer games. :slight_smile:

Don’t get me wrong!
I don’t see over than 1000 users simultaneously connected in my case.
They can be offline and receive notifications instead realtime stuff.

I mentioned open world, but like you really said, it’s a broad topic.
Was just a metaphor.

My target at this time is an application, not a game.

Reading the documentation and a bunch of topics, I figured out that the solution would be a lobby, using groups as lobby.

This app is something like a UBER/LIFT app.
The drivers are sellers, where each one hold a small amount of tickets in real life.
The vehicles are buses, railway and vessels.
When someone need a ticket, a request is sent to the lobby.
The first seller that catch the buying intention, win it.

Today I can’t disrupt this working scenario.
Where sellers hold the physical tickets.
They are distributed by the owners of vehicles.
It’s a very old fashioned system I know.
But it’s take time to break the paradigm.

It’s a very wide region.
Where a vehicle can easily take 1 month to travel back to the starting point.
The capacity of each vehicle can be 100 to 1500 people.

In real life, is something about 10/100 sellers that spreads over 50 groups. There are 100/10000 buyers by day over each group.

There are seasonal peaks, but the benchmark in docs using CockroachDB is above the expectation.

I just asked for some numbers with postgresql because I am already using it.
It makes my life much more ease.

I really appreciate your answer and time.
Thanks! :grin: