written by John Hugg on September 21, 2010
VoltDB, like many distributed systems, achieves high availability through redundant processing nodes. VoltDB calls this K-Safety. Essentially, the distributed system can answer any request at at least K+1 servers, so it can tolerate at least K hardware failures. The operator specifies the value of K that they find is the best tradeoff between failure, robustness and cost. Other systems use the terms “replica set” to describe similar functionality.
Let’s talk about EC2-style clouds that provide you with a virtualized server at an hourly cost.