<p>Do you need full peer-to-peer clustering?</p> <p>Peer to peer clustering allows any node in a cluster to respond to datastore requests. It also covers the mechanisms of replicating data between nodes and also node discovery. It is very robust compared to more common solutions like <a href="">sharding</a> or read-slaves. However it brings with it a lot of complexity in the construction of a robust store structure, operational issues with provision of storage nodes and potential application complexity.</p> <p>To help answer it might be worth running through a few disaster situations and looking at what you would like the system to do. If you can tolerate less than maximum availability for read and write operations you have a broader range of options open to you. It is also worth considering that it is not worth running a complex distributed management system if you are only going to have a few nodes in the cluster.</p>