How to design a large cluster to cache 1 TB of data?
754801Feb 23 2010 — edited Jul 14 2011I am evaluating Coherence to cache large amount of transactional data in the grid, in the order of 1TB.
- Are there customers who have successfully cached such amount of data for improving performance of reads and writes? What is the configuration of the infrastructure - how many nodes/hosts? What quality and types of hardware are recommended for such grids?
- What are the considerations for using large JVMs (10GB per node)? I understand we will be able allocate large memory per node using partitioned backing and NIO buffers from 3.5. Is this recommended for my cache of 1 TB? What are the sizing considerations for using such heaps?
Please share your inputs on architecture
Regards,
Sairam