Chubbly wrote:
-XX:+UseParallelGC -XX:+UseParallelOldGC -XX:-UseGCOverheadLimit -XX:NewSize=512m -XX:MaxNewSize=512m -XX:SurvivorRatio=8 -Xms2048m -Xmx2048m
Does anyone suggest any other ways to tune the JVM? or know the maximum number of strategies or methods for running such a large number?
I'd suggest using the 32-bit Java SERVER VM. I understand that it's marginally quicker than the 64 bit implementation, and also it has much better global native code optimization for high throughput.
You can get it as part of the JDK, and verify using java -server -version.
If you are using the JForex platform as the host for the Strategies, then you must run it also using the Server VM. email me if you need a way to do that on Windows.
My current experience with Dukascopy Remote Server is that performance is really bad, in the sense that onTick latencies are poor. Running it on my local server is both controllable and fast, although I would definitely prefer a Co-Located High Performance system. For me, it doesn't look like Dukascopy's Remote Servers are fast enough, although their network latencies would be much better.
HyperScalper