Device Polling on FreeBSD
As mentioned in my previous post, the Device Polling makes a HUGE difference.
Here is the "before" graph

And here is the "after"

To be fair, the "before" one was taken in the weekend, where we had about 100 more connections. However, notice how now the usage is at 1/4 of that (with about 570 peak connections). The actual server process took about 6.5% of the CPU with 600 connections.
And as a clarification, the spikes you see are not caused by the game server but by various stuff such as recompiling the test server, stress testing, etc.
I am so happy with FreeBSD that I donated 100 USD to the FreeBSD Foundation. I also donated 50 USD to the OpenBSD guys a few months ago, for their effort.
Here is the "before" graph

And here is the "after"

To be fair, the "before" one was taken in the weekend, where we had about 100 more connections. However, notice how now the usage is at 1/4 of that (with about 570 peak connections). The actual server process took about 6.5% of the CPU with 600 connections.
And as a clarification, the spikes you see are not caused by the game server but by various stuff such as recompiling the test server, stress testing, etc.
I am so happy with FreeBSD that I donated 100 USD to the FreeBSD Foundation. I also donated 50 USD to the OpenBSD guys a few months ago, for their effort.
0 Comments:
Post a Comment
<< Home