"If a machine dies within a cluster, we automatically reorganize the cluster to avoid a single point of failure," said Purdy. "We don't lose transactions if one machine goes down, and then another.

"Similarly, we can detect events inside a server such as an infinite loop. A lot of systems are up 24x7 and don't get taken down for maintenance. The end result is that our Guardian technology help us support true continuously available systems," he added. "Within a node, the threads are now responsible for watching each other and ensuring that the node is healthy."

Sometimes a pool of data, such as authentication data for a single sign on environment, needs to be shared by several applications. "Coherence*Web allows applications to access the infrastructure from outside the cluster, yet have isolation between where we manage session data and where it is being used," said Purdy. "Isolation is important for security or to make sure that different parts of systems remain independent."