A decentralized framework for cross administrative domain data sharing

Federation of messaging and storage platforms located in remote datacenters is an essential functionality to share data among geographically distributed platforms. When systems are administered by the same owner data replication reduces data access latency bringing data closer to applications and enables fault tolerance to face disaster recovery of an entire location. When storage platforms are administered by different owners data replication across different administrative domains is essential for enterprise application data integration. Contents and services managed by different software platforms need to be integrated to provide richer contents and services. Clients may need to share subsets of data in order to enable collaborative analysis and service integration. Platforms usually include proprietary federation functionalities and specific APIs to let external software and platforms access their internal data. These different techniques may not be applicable to all environments and networks due to security and technological restrictions. Moreover the federation of dispersed nodes under a decentralized administration scheme is still a research issue. This thesis is a contribution along this research direction as it introduces and describes a framework, called “WideGroups”, directed towards the creation and the management of an automatic federation and integration of widely dispersed platform nodes. It is based on groups to exchange messages among distributed applications located in different remote datacenters. Groups are created and managed using client side programmatic configuration without touching servers. WideGroups enables the extension of the software platform services to nodes belonging to different administrative domains in a wide area network environment. It lets different nodes form ad-hoc overlay networks on-the-fly depending on message destinations located in distinct administrative domains. It supports multiple dynamic overlay networks based on message groups, dynamic discovery of nodes and automatic setup of overlay networks among nodes with no server-side configuration. I designed and implemented platform connectors to integrate the framework as the federation module of Message Oriented Middleware and Key Value Store platforms, which are among the most widespread paradigms supporting data sharing in distributed systems.