Coherence - memory per node - Best practices
499833Sep 23 2010 — edited Oct 1 2010Hi,
Have a question to the coherence gurus.
We have a 200 nodes cluster which stores Rolling last 3 hours data. So as you can imagine we have set expiry of the caches to 3 hours.
So we do create a lot of garbage as well constantly.
Currently if you look at one node its about 55% full. After a Full GC it comes down to 55%.
What is happening is that Nodes are going into frequent Full GC's and at times when too many nodes go for full GC at the same time we are having slowness in the cache.
So my question is what should be the best practice here? is Having the nodes 55% full a bad thing? Will it be able to scale up to 75% and sustain frequent GC's ?
thanks,
Sandeep.