Home

Misc

Presentations

Publications

Students

Department of Informatics

Help

Search

Ongoing and available Master thesis projects

If any of the explicitly mentioned projects looks interesting - or maybe you have ideas for a related project, contact me!

Available Master student projects

• An emulator for a new device in Qemu
  The focus of this thesis will be I/O virtualization. Qemu is chosen as platform because it currently is the only platform to support a number of advanced features that is a prerequisite for this work. The practical part of the thesis will be to implement an emulator for a well-known PCIe I/O device not currently emulated by Qemu, and qualify this work by submitting it to the Qemu open source project. Intel provides publicly available specifications for a lot of their devices which can form the basis for this task. The theoretical part of the thesis will look at aspects of I/O virtualization in light of the implementation made. A suitable candidate would have good C and Linux knowledge and some level of hardware understanding and interest in low level details.

Ongoing Master student projects

• Jan Magnus Opsahl writes a thesis on open source virtualization solutions such as Qemu/KVM and Xen. Virtualization is not a new concept, but has become increasingly important in later years. The introduction of hardware support in recent Intel and AMD processors has made such solutions even more interesting. In addition to experiment with and get a good feeling of some of these tools and features, the thesis will also look at different aspects of recent developments in this area.
  
  

Some finished master theses

Paradial was a Norwegian company that provided solutions for seamless connection in particular for realtime media like video and audio for live communication (voice/video over IP) across any type and number of firewalls. Problems dealt with was both classical network protocol issues with UDP and TCP, newer general purpose protocols such as SIP and H323, as well as relating to a plethora of audio and video codecs and other application layer protocols, such as instant messaging protocols and application sharing protocols. Examples of network problems were: working around unwanted TCP retransmission of expired data, handling and compensating for opaque delay and packet loss patterns in parts of the network not controlled by us. Paradial was acquired by Logitech in spring'2010. FAST was a company that developed advance, state-of-the-art technology for accessing information. FAST developed search engine solutions for existing data and alert systems for future data. FASTs customers included many of the largest companies in the world, and FASTs search solutions were used in projects such as powering yellow pages, simplifying financial analytics, enhancing web shopping sites, intranet searching or aiding prosecuters in crime investigation. FAST was acquired by Microsoft in 2007. Scali developed scalable cluster solutions, very efficient parallel programming libraries and management tools. Customers were mainly companies and institutions with large or many scientific computing problems to solve. Scali was later aquired by Platform Computing.

• Sonam Rani wrote about integrating a SIP based phone service with Facebook. A problem for innovative phone solutions that are independent of the old phone based telephone system is the lack of proper directories to locate the callee and make sure that that person's application is listening for connections. In her thesis she looked at various approaches by using APIs made available by Facebook.
  
  
• Fredrik Oterholt wrote a thesis on Peer-to-peer SIP. SIP is fast becoming a leading standard for managing and establishing complex networked sessions for applications such as among others telephony, video conferencing and network based games. Standard SIP relies on a sentral server, whereas proposed extensions to the SIP standard to support Peer-to-peer signalling allows the server role to be distributed.
  
  
• Nils Petter Rasch-Olsen has written a thesis on search in geographical data. The goal of geographical search is to be able to find documents that matches reuqests like German cities close to the Rhine or Rivers in south France. His implementation was done within the EU project Diligent, where Fast is one of the participants. Nils Petter proposed and implemented a modified version of the R-tree with improved characteristics in search engines where both inexpensive lookup and fast indexing is necessary.
  
  
• Eivind Hasle Amundsen has written a thesis on application of open source projects to implement enterprise search solutions. Currently, companies as FAST, Automony and others serve the business market for search with platforms for search and search related applications. This is opposed to end user search as provided by Google/Yahoo which is very focused on and tuned for the web and openly accessible information. An enterprise search solution is typically a strategic IT choice for a business and may have a big impact on a business' bottom line. Eivinds thesis discusses how and to what extent open source as of today can be used to replace any or parts of an enterprise search solution.
  
  
• Kim Sand has written a thesis on question answering. Search engines allow users to enter a set of query words, and those words are matched with the search engine's index and documents that have the best set of occurrences of those words will be returned as the topmost hits. Question answering is a research field where attempts are made to capture more meaning behind typical natural language queries and answer that instead of just matching the words. This type of matching relies upon a thorough analysis of the text in advance. Kim used Fast technology and plugin interfaces as the framework for the implementation.
  
  
• Vikramjeet Jassal has written a thesis on improved methods for adaptive crawling. A crawler's function is to crawl around a configurable set of pages, such as for instance everything reachable from a certain URL within a set of specified domains, download and feed the content of those pages into the search system. As pages changes, they will have to be re-crawled. One problem is that some pages change more often than others, earlier work [1,2] used statistical methods to determine the rate for re-crawling. Other pages appears (by looking at headers) to change all the time (such as dynamic pages) or because headers are wrong.
  
  
• Håkon Clausen has together with Åsmund Lie written a plugin that uses Fast technology as an index access method of the PostgreSQL relational database engine. Traditional relational database systems are excellent for storage and query of highly structured information. However, when it comes to searching in less strictly structured information such as for example XML documents, relational database systems cannot keep up with modern search engine technology. Focus of Håkons thesis has been challenges in integrating the query languages and how to provide transaction support across the two systems.
  
  
• Åsmund Lie has together with Håkon Clausen written a plugin that uses Fast technology as an index access method of the PostgreSQL relational database engine. Åsmunds thesis looked at generalizing the text index support of the initial plugin implementation to support more complex relational schemas automatically and semi-automatically.
  
  
• Kristian Frøhlich Hansen has written a thesis that explores ways of making IRC content according to user specific needs more easily accessible to the users. Messages and groups are gathered from IRC and feeded through a system that makes it available for search. The work makes use of technology developed by Fast Search and Transfer.
  
  
• Mari Wang has written a thesis on recommender systems for Usenet News messages. News feed implies a large amount of data with a relatively low percentage of sometimes very valuable information for a particular user. Focus is to look at ways of using this information explicitly or implicitly to aid users in finding what they look for. This work makes use of technology developed by FAST.
  
  
• Bjørn Remseth has written a thesis exploring ways to extract information such as preferences, knowledge about earlier choices and other traces of use from a user's files to create an interest profile that can aid that user in subsequent information gathering. You can read his english summary here
  
  
• John Vollestad wrote a thesis on distributed files systems for clusters. An important purpose of the thesis is to evaluate different solutions and ideas for general storage systems for clusters. This thesis is related to my previous work at Scali.
  
  
• Karim Ghouas wrote a thesis on use of the communication protocol VIA over SCI (Scalable Coherent Interface). VIA was designed with high bandwidth-low latency devices in mind, Some results from the thesis are available in english in the paper Consequences of a Zero Copy Implementation, and Comparison with VIA over Myrinet(.ps). This thesis is related to my previous work at Scali.