2006-10-06: DataFed - more distributed system

From Datafedwiki

Back to DataFed Development Events

This is a note on possible ways to make DataFed a fully distributed data system

• Currently, DataFed is a federated data system: the data sources are distributed
• However, the homogenisation, processing and delivery applications are all within the single centrlized DataFed server.
• Hence, DataFed is only partially distributed system

In fully Service Oriented Architecture, most if not all the components are ditributed:

• The metadata for data finding and accessing is distirbuted among the various cataloging systems
• The data are distributed among the data providers/custodians/mediators
• The data access, processing and other are also distributed

The overall system is therefore composed of autonomous components in a NETWORK configuration. In such a system there is also a distribution of

• Networking responsibilities
• Networking benefits and
• Driving forces that make the network connection and flows.

In order for DataFed to evolve, it needs to turn into a networked system composed of autonomous 'DataFed Nodes' wich can provide catalog, data access or processing or other services. These nodes can operate as a network via the common "DataFed Networking Framework" which an perform the various coordination activities..

everything is a service