Killing the server is fine I think. On Mon, Jan 23, 2012 at 9:58 PM, Ali Lown <a...@lown.me.uk> wrote:
> For those that are actually running servers with users on, how is your > server's startup time? > > I have now reached ~60 waves with 30+ blips a message across ~10 > users, and startup time is already ~2 minutes on a small dual-core > (2GHz Opteron) VPS with 1GB of RAM. > > A look at htop/iotop suggests it to mostly be a cpu-processing issue, > though disk access is a constant ~5MB/s while loading. > > I know that a better search implementation (any status on this yet?) (The > pattch that refactors the WaveMap and separates it form SearchProvider is > still under review) > will prevent the need to cache all waves on load. > > Just wondering if any of you have found a way to speed this up a bit? > (@Yuri: WIAB.net must takes hours to start the wave-server?) > waveinabox.net loads about 5 minutes (It runs on VM with 4 virtual CPUs and 7.8 GB of memory/ ~ 6K waves). I never measured what is the bottleneck here, but I assumed it's the IO operations and not CPU. If the issue is with CPU - we can try to process every wave independently in a thread. But that might further complicate the waves loading logic. > > Ali > > PS. Is there a better way to stop a running server than just kill'ing it? > Don't think so. By the way, if you want to run it as daemon - http://groups.google.com/group/wave-protocol/browse_thread/thread/d9cdcea0d5393eda