Projects

• Masters Student Projects

  The following list contains masters student projects which I am willing to supervise. If you are interested in any of these projects, or have an idea for a project which is in the areas I am working in please email me to setup a time to discuss the project.

  • Distributed Context Computing

    A previous project involved the development of a context framework for Android powered mobile phones. This system is able to register various sensors and evaluation complex condition expressions over the values of the data collected from the sensors. Thus it is possible to register a callback when the context of the phone is in the requested state. In this project the student will extend this framework to allow for distributed expressions. That is the framework will be extended to support expressions which evaluate to true when the context on multiple devices satisfies the expression. The distributed communication will of course be handled via our Ibis communication library. Open questions include the best way to represent distributed expression, the security implications of expression registration and result callbacks. The student is expected to build an application which demonstrates the power and functionality of the resulting framework.

  • Energy Aware Optimizations for Context Collection

    A previous project involved the development of a context framework for Android powered mobile phones. This system is able to register various sensors and evaluation complex condition expressions over the values of the data collected from the sensors. Thus it is possible to register a callback when the context of the phone is in the requested state. In this project the student will extend this framework to rewrite the terms of a given expression in such a way as to minimize the total energy used by the framework for monitoring a given expression. This will involve rewriting a given expression into disjunctive normal form and determining the cheapest energy for a given sensor, and then reordering the expression and using short cut evaluation.This may also involve taking advantage of particular knowledge of a given sensor in its relation to other sensors, for example using the acceleration sensor to estimate a new location instead of using the GPS sensor and then turning the GPS sensor back on when the estimate is close to some desired location. The student may also measure the usage of various sensors in order to determine the relative energy consumption, and adapt the framework in order to turn on or off various sensors in order to further optimize total energy consumption. The student is expected to benchmark the resulting improvements of the framework to show the reduction in energy usage for various expressions.

  • Intelligent Computation Offloading

    A previous project developed a computational offloading framework. This framework allows application programmers to easily offload portions of the computation for their application to cloud resources based on available communication and computational resources. This project will add intelligence to the offloading process using profiling and estimation of energy usage, network bandwidth and computation time based on parameters of the input to a given function in order to make smart decisions about when to offload computation. The student is expected to benchmark the changes they make to the framework in order to show that their modified framework is able to make better decisions about computational offloading. Evaluation is expected to cover latency of result, energy consumption, and communication costs among other aspects.

  • Hybrid Distributed Naming For Peer to Peer Overlay Networks

    Truly distributed peer to peer computing on mobile platforms has a distinct problem of naming of resources. Users would like to be able to access resources using a friendly name, however, frequent topology changes as mobile nodes move from WiFi to Cell to a different WiFi network means that addresses associated with names are frequently updated. Centralized solutions are subject to single points of failure, and have scalability issues with many participants submitting many updates as they roam. One solution to the naming problem is to rely on multicast DNS as is used in ZeroConf networking which allows you to discover the names for services within the local network segment. This has the advantage of quick lookup but the disadvantage of not working if there is no participating in the overlay network amongst nodes in the local segment. Another solution often used is that of a Distributed Hash Table (DHT), however DHTs have been shown to have high latencies due to the order log(n) routing inherent to such networks. In this project we propose to build a hybrid system which uses ZeroConf networking and local caches combined with a backing DHT in order to have low latency name resolution in cases where nodes are participating in a single segment with fallback to DHT based naming if no name is found in the caches of local nodes. The student is expected to explore the trade offs with respect to cache coherence and latency of various techniques available within this design space and provide benchmark data exploring the properties of this hybrid naming system.

• Bachelor Student Projects

  I am open to supervising bachelor student projects related to my areas of interest. If you are interested in working on mobile phones or with context or data management please email me to setup a time to discuss a project topic.