Abstract :
Management of semi-persistent state, such as user-session state, is one factor that complicates failure management in clustered three-tier Internet applications (Jacobs, 2003). We observe that the specific properties of user-session state can be exploited to design a lightweight state storage layer that offers many of the same ease-of-management and ease-of recovery properties as stateless components such as Web servers. We describe SSM, a self-tuning, self-protecting, and self-healing session state management layer that provides a storage and retrieval mechanism for semi-persistent, serial-access user session state. SSM is fast, scalable, fault-tolerant, and recovers instantly from individual node failures. Any SSM node may be rebooted at any time and there is no special recovery code, so the performance cost of "eager" recovery is near zero, simplifying recovery policy management when SSM is integrated into a larger system.
Keywords :
Internet; fault tolerant computing; middleware; persistent objects; software fault tolerance; SSM; Web server; eager recovery; ease-of recovery; ease-of-management; failure management; fault-tolerant system; node failure recovery; performance cost; recovery policy management; retrieval mechanism; scalable system; self-healing system; self-protecting system; self-tuning system; semipersistent state; serial-access user; session state management layer; state storage layer; stateless component; storage mechanism; three-tier Internet application; user-session state; Admission control; Aggregates; Availability; Conferences; Costs; Fault tolerance; Internet; Redundancy; Scalability; Web server;
