Data streaming algorithms and compressive sensing techniques, developed originally in theoretical computer science and image-processing contexts respectively, have recently found many applications in computer networking. Single-node data streaming algorithms have been developed to extract important summary information or statistics, such as entropy, flow size distributions, and flow counts, from a single stream of packets passing through a high-speed network link, using a small amount of high-speed memory. Multi-node data streaming algorithms are proposed to extract such summary information from the union of multiple packet streams without the need to physically aggregate these streams together, which may incur prohibitive communication cost. Compressive sensing techniques have been employed to detect sparse (small in the number of affected nodes/links at any moment of time) network events such as sudden dramatic change in traffic volumes and the existence of dirty network measurements, and to recover network statistics vectors (e.g., traffic matrices and flow sizes) that are sparse (containing few large scalars) in nature. Data streaming and compressive sensing are closely related since random projection is the underlying methodology for both.
While both topics have been featured in earlier DIMACS workshops within other special focus series, the majority of the work there do not directly address research challenges faced by the networking community. In this workshop, we aim at bringing together researchers in both networking and theory communities who are interested in solving real-world networking problems using new or existing data streaming and compressive sensing techniques. The workshop will invite/feature such work not only in the more established application domain of Internet measurements, but also in other emerging contexts such as wireless and sensor networks.
We also plan to propose a JSAC (Journal on Selected Areas in Communications) special issue on this topic to JSAC editorial board and serve as guest editors of the issue.