Bachelor and Master Theses

List of theses available for opponentship



List of theses available for opponentship

For the presentations on January 25 and 26, 2018

DVA428

OPTIMIZING ENERGY CONSUMPTION OF REAL-TIME CLOUD COMPUTING SYSTEMS

Ali Alshiekh

Advisor: Hamid Reza Faragardi and Mohammad Ashjaei

Examiner: Thomas Nolte


Abstract:

The huge number of network devices in cloud data-centers consume copious amounts of power, this has emerged as a matter of concern for data center operators. Cloud computing nds its application in various elds, such as: telecommunication, multimedia etc., most of these applications require a timing guarantee for its provided services (real-time applications). Such a cloud computing system with real time applications is often called a Real-Time Cloud (RTC). Once cloud computing is going to guarantees performance and quality of service requirements such as deadlines, more hardware resources are required such as more number of servers, switches, higher network bandwidth, more cooling and power distribution systems. In this context, increasing hardware de- vices and links lead to boost power consumption and augment operational cost. This research work aims at optimizing the network energy consumption, which in turn leads to reduction in energy consumption of data-centers, while satisfying real time constraints. The energy reduction of net- work achieved by turning o idle switches and Playing with the allocation of the workloads to servers to minimize the communication load, thereby increasing the number of idle switches. Our implementation exhibits great-full performance of using OSPF routing mechanism and applying Elastic-tree mechanism over OSPF tasks paths, to give improvement 50% at least of optimizing energy consumption overall network equipment.


Link to the thesis:

http://www.idt.mdh.se/examensarbete/index.php?choice=show&id=2153

*******************************************************************************


DVA501

Effectiveness Of Fault Prediction

Albi Dode

Advisor: Adnan Causevic

Examiner: Jan Carlson


Abstract:

The research community in software engineering is trying to find a way on how to achieve the goal of having a fault-free software. The industry that will use a near fault free software will have it easier to lower the costs of maintenance and the versions of delivered software will be more qualitative. In this case, fault prediction can be used in order to achieve the above objectives. Fully applied fault prediction is not yet achieved on an industrial scale. There is some progress attained in the field during recent years. But knowing and understanding what available tools and algorithms regarding fault prediction can give is yet a goal to be achieved by the industry. In this thesis, several fault prediction algorithms and metrics combinations are tested in an industrial and open source project. The findings helped to understand how the results can vary according to the project’s nature. The main goal is to understand how much fault prediction is integrated and effective in a continuous delivery environment using real case scenarios. The manually collected data, from several versions and in different time periods were applied using two already present algorithms: Nave Bayes and Clustering. As a result, while the usage of this model depends on the company needs, further research in the field can be extended.


Link to the thesis:

http://www.idt.mdh.se/examensarbete/index.php?choice=show&lang=en&id=2053

*******************************************************************************




  • Mälardalen University |
  • Box 883 |
  • 721 23 Västerås/Eskilstuna |
  • 021-101300, 016-153600 |
  • webmaster |
  • Latest update: 2017.10.14