Server Inconsistent latency as playercount increases

[tommy2bar]

I've noticed that the ping on the scoreboard increases steadily as the playercount increases on the server, though an ICMP ping I'm running in commandline to the server does not.
At 0 players I'm getting a ping of 48ms on the scoreboard, which lines up with the ping in command prompt, then at 64 players the scoreboard displays 100-120ms while the ping command is still sitting around 48ms.

I'm new to running servers, pretty much using an RS2 server on a Windows VM to dip my toes into AWS (have to learn it for work).
I'm running the server on a t2.medium EC2 instance which has 2 cores and 4GB of ram, although at 64 players the CPU usage only hovers around 50% and RAM usage hovers around 2GB. I'd seen this issue on other people's servers and assumed it was performance related but the VM really doesn't seem to be getting hammered all that much. Could it be a NIC config thing or some Windows networking configuration issues?

Or is the RS2 server more reliant on single threaded performance? t2.medium EC2 instances only run at 2.4GHz on both cores, I might fire up a C4 or C5 instance which has higher clockspeeds (up to 3GHz) to see if it helps.

 

[Rivimies]

Or is the RS2 server more reliant on single threaded performance? t2.medium EC2 instances only run at 2.4GHz on both cores, I might fire up a C4 or C5 instance which has higher clockspeeds (up to 3GHz) to see if it helps.

Yes, absolutely. RS2 server performance is determined solely by single thread performance. Server is not single-threaded per se, but main thread is the only thing that matters.

I might fire up a C4 or C5 instance which has higher clockspeeds (up to 3GHz) to see if it helps.

3GHz helps a little, but unfortunately it's not enough to run a smooth 64 player server. I would say it will do 48 players fine. Please note that maps vary, some are stressing server CPU more than others.

Could it be a NIC config thing

You can configure NIC to perform better with game server type usage (lots of small outgoing packets) but that is a minor thing. Perhaps 2% performance increase (my wild guess).

I have a longish history of hosting RO2 and RS2 servers (Divided servers for RO2 and rs2.fi servers for RS2). If someone needs a affordable RS2 server in Europe, you can ask me. It's just like renting one from GSPs. I can not promise 24/7 support but otoh I know how these servers work the best.

 

[tommy2bar]

Ah that clears things up for me, thanks for the response!

TWI / Antimatter should really look into optimizing / threading the server since most cloud providers are on the low end of the frequency spectrum on CPUs. Hope AMD's Epyc / Epyc 2nd gen bring server clockspeeds up, Intel's HCC chips are all hovering around 2.4 - 3.0 GHz.

I have a longish history of hosting RO2 and RS2 servers (Divided servers for RO2 and rs2.fi servers for RS2). If someone needs a affordable RS2 server in Europe, you can ask me. It's just like renting one from GSPs. I can not promise 24/7 support but otoh I know how these servers work the best.

Thanks for the offer, I mainly set up my sever to have a poke around at AWS though so I don't mind dropping the player cap a bit.
If I had a decent fiber connection at home I could probably run it on my spare PC and get better performance than AWS... LOL!