Abstraction Techniques for Distributed Massive Event Systems

  • D'Hondt, Theo (Administrative Promotor)
  • Lombide Carreton, Andoni, (Mandate)

Project Details

Description

Problem description:
Computer hardware is becoming increasingly cheap, portable and energy-efficient. Furthermore, more and more such mobile devices are able to communicate with each other via wireless connections. Because of the proliferation of these devices, we see that computational power becomes ubiquitous in our everyday life. This observation has led to research areas such as pervasive computing, where mobile devices communicate with each other in a unobtrusive and dynamic way to support their users in their everyday tasks. Existing programming abstractions that have been developed for this situation cannot be used when these wireless networks take the form of massive, dynamic clouds of volatile information. This is a realistic vision of the future when looking at the miniaturization and cheap mass production of computer hardware, such as RFID tags. Existing techniques assume a limited number of enumerable communication partners over which can be iterated and do not reflect the strong dynamics of the mobile network. Furthermore, interaction with the network has to happen in an event-driven way. Here the problem is that using event-driven architectures introduce a great amount of complexity, which is even worse when looking at massive event systems.

Goal:
The goal is to research programming language abstractions that should allow application programmers to work with the massively generated events on a conceptual level and to couple the event-driven architecture to sequential application components that run on classical computer hardware. Since the cloud of information surrounding a mobile device can be so large and dynamic, it cannot always be divided in individual components. Hence, it is unfeasible to represent this cloud in a classic data structure such as a vector or event queue because such data structures are only a snapshot of the state of the network. We are going to look in a suitable abstraction to represent such a dynamic cloud of information and a way to specify operations on such an abstract pool of data.
The ambient-oriented programming paradigm is specifically aimed at wireless, mobile networks and pervasive computing. This paradigm already assumes support and abstraction techniques for asynchronous communication in mobile networks, which implies an event-driven architecture. We are going to try to extend this paradigm with abstractions that allow programmers to cope with a massive number of events and with massively present but computationally restricted network entities such as RFID tags. This extension should be unified as consistently as possible with the well-known object-oriented paradigm. This way, our goal is to bridge the gab between the distributed, inherently event-driven outside of an ambient intelligent application and the local, object-oriented inside of such an application.

The research will be conducted at the Programming Technology Lab (PROG) of the Vrije Universiteit Brussel. Within this lab there is already vast experience in the area of ambient intelligence, event-driven systems and software and language engineering in the broad sense of the word.
AcronymIWT393
StatusFinished
Effective start/end date1/01/0831/12/11

Keywords

  • mobile networks
  • informatics
  • Pervasive and ubiquitous computing
  • RFID tags
  • event-driven systems

Flemish discipline codes

  • Mathematical sciences