Seminars at the Faculty of Informatics

Liquid Stream Processing on the Web: a JavaScript Framework

You are cordially invited to attend the PhD Dissertation Defense of Masiar BABAZADEH on Wednesday, October 25th 2017 at 15h30 in room SI-008 (Informatics building)

The Web is rapidly becoming a mature platform to host distributed applications. Pervasive computing application running on the Web are now common in the era of the Web of Things, which has made it increasingly simple to integrate sensors and microcontrollers in our everyday life. Such devices are of great interest to Makers with basic Web development skills. With them, Makers are able to build small smart stream processing applications with sensors and actuators without spending a fortune and without knowing much about the technologies they use. Thanks to ongoing Web technology trends enabling real-time peer-to- peer communication between Web-enabled devices, Web browsers and server-side JavaScript runtimes, developers are able to implement pervasive Web applications using a single programming language. These can take advantage of direct and continuous communication channels going beyond what was possible in the early stages of the Web to push data in real-time.

Despite these recent advances, building stream processing applications on the Web of Things remains a challenging task. On the one hand, Web-enabled devices of different nature still have to communicate with different protocols. On the other hand, dealing with a dynamic, heterogeneous, and volatile environment like the Web requires developers to face issues like disconnections, unpredictable workload fluctuations, and device overload.

To help developers deal with such issues, in this dissertation we present the Web Liquid Streams (WLS) framework, a novel streaming framework for JavaScript. Developers implement streaming operators written in JavaScript and may interactively and dynamically define a streaming topology. The framework takes care of deploying the user-defined operators on the available devices and connecting them using the appropriate data channel, removing the burden of dealing with different deployment environments from the developers. Changes in the semantic of the application and in its execution environment may be applied at runtime without stopping the stream flow.

Like a liquid adapts its shape to the one of its container, the Web Liquid Streams framework makes streaming topologies flow across multiple heterogeneous devices, enabling dynamic operator migration without disrupting the data flow. By constantly monitoring the execution of the topology with a hierarchical controller infrastructure, WLS takes care of parallelising the operator execution across multiple devices in case of bottlenecks and of recovering the execution of the streaming topology in case one or more devices disconnect, by restarting lost operators on other available devices.

Dissertation Committee:

  • Prof. Cesare Pautasso, Università della Svizzera italiana, Switzerland (Research Advisor)
  • Prof. Marc Langheinrich, Università della Svizzera italiana, Switzerland (Internal Member)
  • Prof. Robert Soulé, Università della Svizzera italiana, Switzerland (Internal Member)
  • Prof. Gustavo Alonso, ETH Zurich, Switzerland (External Member)
  • Prof. Tommi Mikkonen, University of Helsinki, Finland (External Member)