big-bang takes 100% CPU. normally this is not a big deal because you have to take the server offline to perform big-bangs. however, now that twgs allows big-bang while other games are active, performing a big-bang effectively lags out the players in other games.

Multi-core CPU, ftw

yeah i don't have a multi-core cpu on my twgs, but even so, have you tried a big-bang with live games and did you confirm that a different core was chosen to do the work? i'm betting that it makes the problem worse than a single core.

say you have 10 players in a single game and 2 cores. 5 players' TW2002 processes are served by core A, and the other 5 by core B. then you perform a big-bang on a new game. core B is chosen by the OS to do the big-bang, and is pegged at 100% for 90 seconds. for those 90 seconds, 5 players on core B are lagged out, but the other 5 players (on core A) are not. 5 players have a serious advantage for 90 seconds. call it luck of the draw, but i guarantee you get complaints if players suffer in-game consequences.

Not currently running the beta, but it's relatively easy to
peg the CPU at a 100% on one node. Hyper-threading
and multi-core setups make a huge different there. It's
very noticeable.



Right, it schedules the core with the lowest load first,
generally. So it would pick the lowest used core, and
bang on that. For those 90 seconds, the players on
core B would be lagged. The other 5 players would be
better off, but not perfectly fine. Any IO will still leak
across to other cores. I'm skeptical you'd see a
major advantage.

The players could always log out and back in to get
a new core if it was a serious problem. Or maybe
there's a way to re-assign the threads as they're
running.

When I was rebanging my test on ctw, I was seeing
some CPU spikes, but not 100% for 90 seconds.

Maybe there's a way to speed up the bang, or spread
it out over multiple threads. Dunno.

It's seems the majority of the CPU is utilized during the max course length process. Maybe a way to slow down the big bang process some might help. It would take longer for big bang to complete, but possibly effect the active games less.

Tested on my laptop:
AMD Phenom(tm) II N950 Quad-Core Processor 2.10 Ghz
I will retest on the server in the morning it is a AMD Phenom II Quad 3.2

Searching for Unaccessable Clusters 35%

Setting Max Course 33%

Yes, that's about what mine uses, 20% to 40%, but the players do notice that when I'm banging a game they get considerable lag in the games for the duration of the big bang process especially during the last few tasks.

I rarely ever rebang on the server for that reason, I usually build the bangs on laptop and move them over. I have a lot of changes I make prior to opening so it is easier to do them on laptop.

if you have a quad-core machine, then 25% utilization means one entire core is at 100%. that's how it's reported. so in toyman's case (35%, quad-core), 100% of one core and 40% of another core is being used during bigbang. not sure how many cores bigd has, but 20-40% utilization is a lot for a multi-core machine.

I watched the 4 bar graphs and the first core never maxed it went to about 60% then the second core went up slightly. Dunno with Microsoft though 66% on the graph could = 100%
I re ran and took a few snap shots, the Course Length has more info. I hope this helps. I will try it shortly on the server.
Laptop Running Win7

Server CPU Usage for Big Bang:
Server Running on WinXP Pro

Have any of you treid disableing the Hyperthreading? You may get better performance out of twgs v2 as all cpu's would be timed equeally by the OS and not max cpu 1 before the others as when hyperthreading is enabled. I know it sounds crazy but... you may want to try it.

interesting stuff. my twgs is an amd k7 1ghz, 256 mb ram. historically, that was more than enough. it seems like other sysops are throwing more hardware at it than me. toyman's big-bang showed ~70% utilization of a single core during big-bang. it's no wonder that big-bang operations required a complete shutdown of the twgs in the past. now that we're able to do big-bang while hosting games simultaneously, perhaps better hardware is warranted.

I was in the need of a new server, and it was only $400 to go with a Asus MB, Quad 3.2, 4 Gig of Ram and a New P/S. Was only $75 ~ $100 to scale down, figured what's the point I might use it for something else one day. Currently with server side scripts running and 10 people moving it rarely goes over 3% CPU during normal play. And that is with me VNC'd in since I rarely have a monitor hooked up to it.

I completely missed this thread, but I recognized an issue here myself while banging games on twgs.classictw.com during active play. A few weeks ago, I went in and added a setting (under Global Settings in Command Center) to specify how much CPU Bigbang should use. If you have a single core and set to 50% CPU, then Bigbang only pulls 50% load on the CPU. If you have dual-core or two CPUs, Bigbang runs in a single thread, so it'll pull 25% load. So if you want to let your players play through a bang and don't mind that it'll take awhile, you can set your percentage as low as you feel is needed. Sometimes a lengthy background bang is better than a quick offline bang.