An embedded co-processor for accelerating window joins over uncertain data streams

摘要

In many new applications fields, such as surveillance, networks, and sensor technologies, the generating data are time-varying, unpredictable, uncertain, continuously arriving, and must be processed online. To keep up with these rapid data streams, accelerating data processing is one of the key problems in a UDSMS (Uncertain Data Stream Management System). Most of the algorithms that are proposed in the literature are implemented in software and can potentially result in CPU saturation. To greatly promote UDSMS processing speed, we focus on accelerating join processing, which is one of the most costly operations. We propose, for the first time, a hardware co-processor called UWJSP (Uncertain data Window Join Special co-Processor) for implementation. UWJSP is composed of several processing units. Each unit executes special instructions that continuously track changing queries. We also propose an integrated method for improving the scalability and flexibility. The experimental results reveal that UWJSP can achieve an order of magnitude improvement over a software implementation. The low cost and high performance UWJSP can be used in applications where real-time performance is indispensable.