Apr 19, 2007: If You Build It They Will Come: The snBench Architecture
Filed in: Colloquium
Dr. Azer Bestavros, Boston University
The power of infrastructure-less sensing-oriented networks lies in their refinement and optimization to achieve a singular task or goal. By contrast, the power of emerging networks of sensors embedded in common, public spaces (which we call Sensoria) lies in their flexibility and programmability. Toward this end, we propose the snBench, an application development, deployment, and execution architecture oriented toward enabling accessible programming, efficient rollout, and extensible Sensorium component capabilities. We will present an overview of the snBench architecture, highlight its research challenges and the research topics it catalyzes, and discuss (and demo) our first-generation implementation of a Sensorium of Video Sensor Networks.
The components of snBench are most easily thought of in terms of analogous components in traditional stand-alone development environment. SNAFU (SNet Applications as FUnctions) is a high-level programming language (analogous to Java or Scheme) featuring a functional computation model plus a "trigger" mechanism for embedding and predicating computation in time. SNAFU is then compiled into a DAG-structured virtual instruction set architecture, STEP (snBench Typed Execution Plan). The Service Dispatcher (linker/scheduler) then takes this STEP program and breaks it apart, linking its constituent sub-graphs to individual nodes running the Sensorium Execution Environments (SXE) and binding those sub-graphs together with appropriate network protocols. The Service Dispatcher can schedule many such programs onto a single Sensorium network simultaneously, taking advantage of programs' shared computation and dependencies to make more efficient use of network, computation, and storage resources.
Time permitting, I will also talk about our effort to allow some of the safety and QoS properties of snBench artifacts to be formally verified (or certified) using the TRAFFIC framework - a novel type system and associated type spaces, which constitute accessible representations of the results and conclusions that are derivable using complex compositional theories.
The snBench work is pursued in collaboration with Michael Ocean and Assaf Kfoury, building on prior work with Adam Bradley (PhD'03, now at Amazon Research). The TRAFFIC work is pursued in collaboration with Likai Liu, Assaf Kfoury, and Ibrahim Matta.
Azer Bestavros obtained his PhD in Computer Science from Harvard University in 1992, and since then has been on the faculty of the CS Department at Boston University, where he is currently Professor and Chairman. His research interests are in the general areas of networking and real-time systems. His seminal works include his generalization of classical rate-monotonic analysis to accommodate probabilistic guarantees, his pioneering of the Internet reverse proxying model adopted years later by CDNs, and his characterization of Web traffic self-similarity and reference locality. His research work has culminated so far in 10 PhD theses, over 80 masters and undergraduate student projects, six patents, and two startup companies. With over 3,000 citations to his papers, CiteSeer ranks him in the top 5% of its list of 10,000 most-cited CS authors. His research has been funded by government and industry grants totaling over $15M. He chairs the IEEE TC on the Internet, received distinguished service awards from both the ACM and the IEEE, and is a distinguished speaker of the IEEE.
Abstract
The power of infrastructure-less sensing-oriented networks lies in their refinement and optimization to achieve a singular task or goal. By contrast, the power of emerging networks of sensors embedded in common, public spaces (which we call Sensoria) lies in their flexibility and programmability. Toward this end, we propose the snBench, an application development, deployment, and execution architecture oriented toward enabling accessible programming, efficient rollout, and extensible Sensorium component capabilities. We will present an overview of the snBench architecture, highlight its research challenges and the research topics it catalyzes, and discuss (and demo) our first-generation implementation of a Sensorium of Video Sensor Networks.
The components of snBench are most easily thought of in terms of analogous components in traditional stand-alone development environment. SNAFU (SNet Applications as FUnctions) is a high-level programming language (analogous to Java or Scheme) featuring a functional computation model plus a "trigger" mechanism for embedding and predicating computation in time. SNAFU is then compiled into a DAG-structured virtual instruction set architecture, STEP (snBench Typed Execution Plan). The Service Dispatcher (linker/scheduler) then takes this STEP program and breaks it apart, linking its constituent sub-graphs to individual nodes running the Sensorium Execution Environments (SXE) and binding those sub-graphs together with appropriate network protocols. The Service Dispatcher can schedule many such programs onto a single Sensorium network simultaneously, taking advantage of programs' shared computation and dependencies to make more efficient use of network, computation, and storage resources.
Time permitting, I will also talk about our effort to allow some of the safety and QoS properties of snBench artifacts to be formally verified (or certified) using the TRAFFIC framework - a novel type system and associated type spaces, which constitute accessible representations of the results and conclusions that are derivable using complex compositional theories.
The snBench work is pursued in collaboration with Michael Ocean and Assaf Kfoury, building on prior work with Adam Bradley (PhD'03, now at Amazon Research). The TRAFFIC work is pursued in collaboration with Likai Liu, Assaf Kfoury, and Ibrahim Matta.
Bio
Azer Bestavros obtained his PhD in Computer Science from Harvard University in 1992, and since then has been on the faculty of the CS Department at Boston University, where he is currently Professor and Chairman. His research interests are in the general areas of networking and real-time systems. His seminal works include his generalization of classical rate-monotonic analysis to accommodate probabilistic guarantees, his pioneering of the Internet reverse proxying model adopted years later by CDNs, and his characterization of Web traffic self-similarity and reference locality. His research work has culminated so far in 10 PhD theses, over 80 masters and undergraduate student projects, six patents, and two startup companies. With over 3,000 citations to his papers, CiteSeer ranks him in the top 5% of its list of 10,000 most-cited CS authors. His research has been funded by government and industry grants totaling over $15M. He chairs the IEEE TC on the Internet, received distinguished service awards from both the ACM and the IEEE, and is a distinguished speaker of the IEEE.