Observation: Dedicated Server requiring more CPU cycles (ver 1.1.4)

[schroeder_lvb]

Since December of last year, NWI issued several Sandstorm patches.

I've been noticing increase in CPU cycles required to host a Linux dedicated server. Back in December I hosted three 8-slot Checkpoint servers on 2-core Xeon VPS. At full capacity (all 3 servers 8/8 players) I had enough cycles left over for me to move and run my old Insurgency server there as well.

Today I noticed running 2 full Sandstorm dedicated servers will peg out both CPU cores when these are at capacity (2 x 8/8 players). This is with standard (default) bot count.

This evening I shut down my 3rd server to keep from passing adverse effects to client-side like lags, stutters or any synching issues. I haven't heard any performance related complaints from my group running release 1.1.4.

I've not noticed any changes to my cloud metrics (same clock speed, bogomips, etc.)

If you are experiencing server-wide problems recently, check your CPU usage. Let me know if you noticed the same (or not!)

NWI - please consider "dedicated server optimization" at some point in the future. There are some serious bugs yet to be stomped out, too!

[schroeder_lvb]

Thanks W - I was a bit reluctant to post this since I find it difficult to take a good measure of Dedi servers as CPU loading depends so much on what the game is doing (what client gamers are doing, which map, how many bots, how many online players -- this last one is the biggie).

But I wanted to put the information out there since I saw few operators reporting 1.1.4 performance problems.

For me if the situation worsens in the future releases the cost-per-server will exceed my threshold. I'll likely go elastic cloud and run them 12/7 instead of 24/7. What fun.