Quote from: opticalcarrier on December 25, 2013, 08:35:31 PM

Quote from: smartwart on December 25, 2013, 08:28:46 PM

Quote from: jl777 on December 25, 2013, 08:20:13 PM

I have a bunch of dedicated servers, but all at one data center. Is there any benefit from deploying the server code on more than one of the servers?

What would the typical CPU load be? These are compute oriented servers, so not much bandwidth for each, maybe 10mbps.

How much would this help?

James

cpu load is negligible.
every node helps to split the load.

it is java memory intensive though. yes! putem up!

I never understood this. On a hallmarked node in AWS with 3.7GB RAM (2GB heap) I'm seeing very good memory performance out of the NRS application; looking at GC charts it really seems like it wouldn't benefit from any more than 1GB or maybe even 768GB heap on a 1GB server. If anyone is truly seeing memory issues at medium/large heaps then I'd be really interested in seeing the GC logs from that - please feel free to link me a pastebin of the GC logs after running a while using GC log settings such as:

-verbose:gc -Xloggc:gc.log -XX:+PrintGCDateStamps -XX:+PrintGCDetails