Critical Friend Model: A Vision Towards Inter-cooperative Grid Communities

Much work is under way within the distributed computing community in order to assign a job to an appropriate resource discovered from a fully decentralized and heterogeneous infrastructure with reasonable cost, such as optimized job responsible time, executing price, etc. However, local resources of individual virtual organizations (VOs) are managed under independent policies and constraints, therefore existing solutions are normally designed for specific scenarios and lack of commonality. In addition, boundaries of different VOs raise extra difficulties on job sharing and collaboration amongst distributed nodes. On the other hand, the obtained knowledge from multinode cooperations is normally discarded, although it in future may lead to intelligent scheduling decision by means of previous collaboration records and experience. Especially, the trust built up according to historical collaboration between nodes from different VOs may overweigh and cross the boundaries of VOs themselves. In this work, the Critical Friend Model (CFM) grid scheduling solution is proposed to bridge the aforementioned gap between decentralized nodes and VOs. The Critical Friend Model is comprised of a set of general workflows and algorithms to make better use of node’s knowledge of the neighborhood derived from historical collaboration, which is kept in the local storage and known as the metadata snapshot. In addition, a set of related grid components are also introduced to give the visible implementation roadmap of the CFM in the near future.