FAQ

###What is the relationship between a rack and a data center? A Dynomite cluster consists of a group of data centers and each data center has a number of racks. Each rack consists of a number of nodes. Note that different racks can have different numbers of nodes. Similarly, different data centers can have different numbers of racks.

###What is a replication factor? In a Dynomite cluster, each rack contains a complete set of data. Replication factor is the number of complete data copies in a data center. In other words, it is the number of racks in a data center.

###What are the currently supported datastore servers? Redis and Memcached.

###Why not just use Cassandra? Cassandra is written in Java and at runtime, GC trigger can result in unexpected latencies in some cases. Cassandra is fast for writing but in very high frequency writes, Cassandra node will get blown up very soon as the compaction can't catch up. Also during a compaction is running on a Cassandra node, read latency is also affected due to the extra CPU and disk loads.

Furthermore, to achieve the same throughputs, and latencies as Dynomite, using Cassandra, you probably need to spend 15X-20X of the cost of a Dynomite cluster.

If you need more consistency, Cassandra at this point provides a better solution. However, Dynomite is going to provide different levels of consistencies very soon. Stay tuned.

###Why not use MC Router? Do you employ a Read-Modify-Write pattern? Do you love Redis and its API (set, list, hash)? Do you want to simplify your cluster management leveraging a peer-to-peer system? Do you want to avoid an extra network hop that McRouter introduces? If you answered yes to these questions, then Dynomite is for you.

###Why not use Redis cluster? Dynomite is a peer-to-peer system. Redis cluster is multi master/slave system. You should choose whichever system works best for you.