Initially, I want to keep costs fairly tight if possible. I plan to use 3 instances of m1.large servers with 200GB of disk space per instance. The application will use up this diskspace within the first 4 months. At which point, I will increase the disk space per node by adding new nodes to the cluster with larger disks and replace the older instances.
The general idea is to keep the disk lean until I need to expand them to 500GB after which I will just add more new nodes to the cluster. P.S. I haven't done any benchmarks, but the system won't be too CPU intensive, so I wanted to expand the capacity of the disks by "doing a rolling upgrade" until I need to add nodes to the cluster to handle the large disk spaces (e.g., 500GB per node max) or CPU becomes an issue. Does this sound like a good solution? -- -Richard L. Burton III @rburton
