i'm looking for the best server configuration for launching a Cassandra cluster on a datacenter. I believe there are hardware settings that lets you get the best out of Cassandra. For example, I/O of network and disk, disk type (SATA, SAS or SSD), memory size, etc.
I did some research and found the use case of Netflix (http://techblog.netflix.com/2011/11/benchmarking-cassandra-scalability-on.html). In this article and Cassandra's own documentation says that the platform has been developed to achieve better performance with SSDs disks. However, as SSDs are much more expensive than SATAs, it's question: the performance difference worth the investment?
Still quoting the Cassandra documentation (http://www.datastax.com/docs/1.2/cluster_architecture/cluster_planning), the ideal would be to have two disks, one for the Commit Log and another for data directories. If I have a SSD disk for the Commit Log and another disk of cheaper technology for data, will i have satisfactory performance or the best option would be two SSD disks?
So folks, has i told before, i'm looking for the best settings to deploy a Cassandra Cluster in a data center on the cloud. Any other experience or observation that i have not commented is welcome too.
Thanks for the assistance :)