Selective recovery in distributed systems
Sarmistha Neogy, Anupam Sinha, Pradip Kumar Das · 2004
Selective recovery of processes during rollback is a well known technique for avoiding unnecessary rollbacks of a large number of processes after a failure is detected in a distributed system. The technique that selects a minimum number of processes for recovery in the present work is simple and though the selection is on-line, it does not add any overhead. A process may however, be directly or even indirectly dependent on the faulty process and hence needs to recover. The initial tracking of dependence of a process on the faulty process is done by checking local variables. For maintaining consistency a process may have to rollback even if it is not directly dependent on the faulty process.