An Essential Design Pattern for Fault-Tolerant Distributed State Sharing

By: N. Islam, M. Devarakonda
Published in: CACM, Oct. 1996
Pages: 65-74
Category: Distributed Systems, Fault-Tolerant Systems

Summary: This pattern creates local views of global data in a distributed system, maintains consistency between the local and global data, detects processor failures, and recovers global state in the event of processor failures.

Pattern: Recoverable Distributor

This pattern creates local views of global data in a distributed system, maintains consistency between the local and global data, detects processor failures, and recovers global state in the event of processor failures.

• This pattern is similar to Observer [Gamma+95], where observers are updated automatically when a single subject changes. Here, global and (possibly) local state is updated on a local state change.
• A local state manager acts as Proxy [Gamma+95] for the global state manager.
• This pattern is similar to Strategy [Gamma+95] in that it provides a uniform interface for a family of protocols.
• The relationship between the state-sharing and fault-tolerant components in this pattern follows Bridge [Gamma+95]. This allows the Recoverable Distributor to mix and match data-consistency and fault-tolerant protocols.