News
DescriptionA component is defined as a unit of software application composition with contractually specified interfaces and explicit context dependencies that can be developed, acquired, added to the system and composed of other independent components, in time and space. Component interfaces determine the operations that a component implements, and the operations it uses from other components during its execution. A distributed component-oriented model is an architecture for defining components and their interactions. It must provide a packaging technology for deploying binary component executables. Moreover, it needs a container framework for injecting life cycle, thus permitting activation and passivation of component instances. Other services include security, transactions, persistence, events, and others. We have designed a new decentralized P2P grid component model which runs on top of an overlay network. We have implemented the majority of traditional component models services, but adapted them to the underlying topology. These include: - A decentralized component location and deployment facility.
ArchitectureThe container is responsible for managing components life cycle, and notifying them about life cycle events such as activation, passivation, or transaction progress. Any component provides event interfaces that the container automatically invokes when particular events occur. Moreover, it provides components uniform access to services such as persistence, security, transactions, and many others. In traditional client-server based component models, the container itself is usually based on a web application server, database server, operating system, etc. These kinds of containers are rather monolithic and consume large amounts of resources thus requiring powerful machines to run on. This philosophy remains in stark contrast with that of P2P, where machines are usually treated as equals, and applications run on them must adapt for each nodes own capacity and limitations. For our component model we have taken into account these considerations, and opted for designing a decentralized lightweight container model. In our case, all of the nodes that belong to the network are containers, and as such, they can house many components. The idea is that any component can be run in any node (except for any restriction), because each node runs a lightweight container. Our containers are fault resilient and autonomous: components are replicated all along the network. If a container fails, surely other containers housing those components will exist in the network. Downloads
Authors
Publications2005
|