Skip to Main Content
Co-allocation is newly developed technology for downloading data files from multiple replicas by establishing multiple connections in parallel, thus improving performance over single-server transfers and helping to alleviate the internet congestion problem. Schemes including our previous the anticipative recursively adjusting mechanism plus (ARAM+) model based on co-allocation strategies and resource broker which provides capabilities such as service allocating, resource discovery, job scheduling, job monitoring and data access to users' application have been proposed. However, the challenges to the resource broker architectures are bottleneck and single point of failure for file transfer. Moreover, the secure intra-grid communication against internal attacks is another important issue. To overcome these problems, we present a new fully decentralized resource broker model with authorization and authentication based on co-allocation method for redundant parallel file transfer in Data Grid. We design and implement a resource broker agent, calling "light broker" to share the workload of resource broker. We also propose a multiparty key agreement protocol to provide a secure resource group communication. The experimental results demonstrate that, our approach achieves dependable performance with various loads of services, broker failures and possible attacks.