Search code examples
cachinggridgain

Is there a difference between starting GridGain nodes on the same local machine vs on different machines?

Given that a local machine has sufficient RAM, is there a difference between starting N GridGain nodes on that one local machine versus starting N GridGain nodes on N distinct machines (each of which, also has sufficient RAM)

I am looking at storing data in a GridGain cluster and then querying it to see the performance. Any tips or things I should look out for when dealing with this?

I am worried that starting more than one node on the same machine would have an affect, as that means latency between nodes is decrease (? this is a speculation). On the other hand, is there an approach to programmatically start and shutdown nodes on remote machines (under the same virtual network)? My desire to start nodes on the same machine stemmed from the fact that there are well documented ways to start and stop nodes, so long as they are on the same machine.

Thanks in advance

Solution

    1. The only reason to start several nodes on one machine is to control JVM GC pauses. By splitting memory between multiple local machines, you can allocate less heap to every JVM and decrease GC pauses. Another way to decrease GC pauses, is to use Offheap Memory.

    2. You can use command-line or GUI Visor management console to automatically start/stop/restart remote nodes.