While the previous example is a very crude demonstration of replicating data across thin data servers there are, no doubt, many places that could use improvement. While it did solve the problem of maintaining data across a cluster, it did not do so in real time. For the end user the problem is solved, but it does not do much to solve the overall complexity of the system.

The Network File Sharing (NFS) pattern aims to add real filesystem-level replication to the web application cluster by centralizing the mount point for the data. We will improve upon the original example by utilizing an underlying virtual machine whose sole purpose is to host an NFS share for the instances that are accessible through the load balancer. This provides a bit of performance increase as the underlying filesystem can be improved with little or no change to the instances that only read and write.

Again, as with all solutions, there are some benefits and concerns to be aware of. While this does...