[infinispan-dev] Infinispan Large Scale support

[Bela Ban]

On 3/18/11 10:35 PM, Dave wrote:
> Bela,
>
>    No VMs. Each node has a 3GB JVM, dual GB Ethernet (bonded).


Using IP Bonding ? Configured in a *load balancing* or *failover* fashion ?


> But there are other things running on the nodes. We are starting them all at once.

If you start 400+ nodes at the same time, there will be merges going on. 
I suggest stagger startup, or at least the first 10 nodes. There are 
also changes to reduce (discovery) traffic in concurrent startup in 
2.12.1. Plus, we should probably set the view bundling timeout higher.

I'll provide you a first draft of udp-largeclusters.xml Monday, let's 
try with that setup then.


-- 
Bela Ban
Lead JGroups / Clustering Team
JBoss