In this talk we will discuss stream program optimizations that eliminate stream program bottlenecks and deploy stream graphs onto shared-memory multicore architectures. We employ a data rate transfer model which permits the computation of closed forms for data rate transfer functions of actors. A combinatorial optimization problem uses these closed forms to maximize the throughput of a stream program. Although this problem is inherently NP-hard, our 2-approximation algorithm achieves results close to the optimal solution. The talk will conclude with challenges and open problems for the deployment of streaming applications in the Cloud.

About the speaker

Bernd Burgstaller is an Associate Professor at the Department of Computer Science at Yonsei University. He completed his PhD in static program analysis at the Vienna University of Technology in 2005 and continued as a postdoc at the University of Sydney until 2007. Before pursuing an academic career, Bernd Burgstaller worked in industry as programmer and software architect for Philips Consumer Electronics, Vienna.