Leveraging Advanced Network Hardware for Streaming Applications
Recent advances in networking, databases, and Internet applications have resulted in several important trends. Advanced network hardware has emerged to ensure the fast and efficient routing of messages. The increasing sizes of data have led to the development of sophisticated one-pass data stream management algorithms by the database community. Network attacks, spam, Distributed Denial of Service attacks, and Internet fraud in general are now recognized as major problems for diverse internet applications. The confluence of these trends motivates the research in this proposal, which intends to solve some of these internet problems using data streams algorithms implemented on advanced network hardware. In particular, we propose to develop data stream algorithms that are specifically designed for network processors units (NPU) that are integrated with content addressable memories. Given the ubiquitous availability of NPUs, the proposed approach will have the potential for use in the on-line analytical processing of message flows as well as in the identification of network attacks and fraud. The proposed research is high risk since it involves the integration of specialized hardware and software components from different vendors, including Intel for the NPU, IDT or NetLogic for the content addressable memories, and Monta Vista and Teja for the operating systems. This integration process is fraught with risks given the need for one-on-one communication with the support personnel of the different vendors to successfully integrate these novel computing platforms. However, the research will have high impact since, if successful, it will lay the foundations for the widespread use of efficient hardware devices throughout the internet for the detection of spam, fraud and DDoS, as well as enable data-centric analysis of network traffic for the networks of the future.