We will be receiving Matteo Nardelli from University of Rome Tor Vergata who will give the following talk on Wed. Oct. 17th at 11am in F302 (Sicile).
Title: QoS-aware Deployment and Adaptation of Data Stream Processing Applications in Geo-distributed Environments
Abstract: Data Stream Processing (DSP) applications are widely used to process big data streams emitted by distributed sensors and devices in a (near) real-time fashion. Since data sources are, in general, geographically distributed (e.g., in IoT scenarios), recently we have also witnessed a paradigm shift with the deployment and execution of DSP applications over distributed Cloud and Fog computing resources. Besides benefits (e.g., reduced application response time), this computing environment introduces new challenges to deal with (e.g., resource heterogeneity, non-negligible network latencies).
In this talk, I explore the challenges of running DSP applications over geo-distributed environments, and I present the key results of my research activity. First, I discuss the initial application deployment, encompassing the placement and replication of the application operators. Then, since DSP applications are long-running and subject to varying workloads, I present strategies for conveniently adapting their deployment at run-time. These solutions aim to adjust the deployment while explicitly considering the reconfiguration costs (e.g., application downtime caused by state migration). I will show some of our results, obtained using real system prototyes, and then try to identify and discuss a few open problems.
Matteo’s personal page: http://www.ce.uniroma2.it/~nardelli/