Loading...
Please wait, while we are loading the content...
Similar Documents
Fault-tolerant distribution and persistence of objects using replication.
| Content Provider | CiteSeerX |
|---|---|
| Author | Goeschka, Karl M. Smeikal, Robert |
| Abstract | Large and complex object-oriented applications demand both transparent distribution and persistence. When faulttolerance is required, improving single component availability is insufficient, because it does not scale well to large systems. Therefore, we provide scalable fault-tolerance for both distribution and persistence by using only a single mechanism: replication. We thereby introduce a component-based architecture, which clearly separates distribution, persistence and replication. The key idea is to set up the system like a conventional distributed system during normal operation, while propagating persistent object data prepares for degraded scenarios. As proof of concept we present the results of a successful real-life implementation: a distributed telecommunication management system, which uses asynchronous primary copy replication as algorithm for propagating the persistent object-data. The result is a flexible, robust and highly available framework for managing object-oriented software in a distributed system. 1. Problem description Transparent distribution and persistence are widely used throughout today’s object-oriented applications and supported by a number of commercial frameworks (SUN’s Enterprise Java Beans, COM+ and IPersist, CORBA Persistent State Service). As long as all parts of the system are available, this is satisfying, since the system setup provides advantages from both, distribution and persistence. This is sufficient for some applications, but others require more fault-tolerance, where fault-tolerance is the ability of a system to continue functioning while a failure is still unrepaired or even undetected [1]. Improving single component availability is insufficient, but rather fault-tolerance needs to be incorporated into the system architecture itself, as quoted in [1]. |
| File Format | |
| Access Restriction | Open |
| Subject Keyword | Asynchronous Primary Copy Replication Persistent Object-data Problem Description Transparent Distribution Commercial Framework Normal Operation Distributed Telecommunication Management System Object-oriented Application Enterprise Java Bean Persistent Object Data Prepares Corba Persistent State Service Degraded Scenario Single Mechanism Object Using Replication Complex Object-oriented Application Component-based Architecture Distributed System Successful Real-life Implementation Scalable Fault-tolerance System Setup Available Framework Object-oriented Software Fault-tolerance Need Key Idea Large System System Architecture Conventional Distributed System Fault-tolerant Distribution Single Component Availability Transparent Distribution |
| Content Type | Text |
