Internships Summer 2017/18
Proposed Project Title:
Performance Analysis of a Framework for Graph Analytics
Big data is the phenomenon that companies and organisations increasingly collect large amounts of data and analyse this data in order to improve their operation and services, and also introduce new services. Graph analytics is one branch of big data that studies graph-structured data. Graph-structured data captures many types of data, including web links, friendship links in social network graphs and physical distance and routes in the real world. Because "big data" inevitably involves large volumes of data, speed of processing big data is quintessential to delivering useful services.
The goal of this project is to study the execution speed of graph analytics. In this project, the student will be introduced to the principles of graph analytics and will learn how these algorithms are structured by means of very simple operations applied repeatedly to the vertices and edges in the graph. The student will learn how these operations are specified in a framework for graph analytics that is developed in the Centre for Data Science and Scalable Computing. The framework optimises the execution speed and leaves the user to focus only on the functional concerns. Moreover, the student will learn the principles of measuring the execution speed of graph analytics. The main tasks in the project are then:
- To port or implement a number of algorithms that are commonly used in graph analytics, such as triangle counting, collaborative filtering using matrix factorisation, K-core decomposition of graphs and determining maximal independent sets. The logic and operation of these algorithms will be explained and blueprints for the implementation are available. Each of these algorithms can be implemented in about 100 lines of code using the framework.
- To measure the performance of graph analytics on a modern, state-of-the-art server using the Intel Xeon Phi Knights Landing (KNL) processor. The performance of graph analytics will be measured as a function of the number of processing cores used (KNL has 256 cores which can independently perform computations), as a function of the choice of cores (using all does not necessarily deliver best performance) and as a function of the configuration of the interconnection of these cores. Moreover, KNL has a “High-Bandwidth Memory” (HBM), a small memory placed in the processor package that can be configured as a transparent cache or as a directly accessible memory. The impact of this configuration on performance will also be evaluated.
This project combines programming with performance analysis in about a 30/70 relationship. The exposure to big data and performance analysis will provide useful skills that will prove useful in a final-year project and delivers excellent employability prospects.
The successful applicant will be supervised by myself (lecturer) and the PhD student who developed the framework.
Interested students are welcome to contact me for further details.
- To learn about graph analytics, a branch of “big data” that studies graph-structured data
- To implement or port graph analytics operations to a state-of-the-art framework for graph analytics
- To perform an extensive performance evaluation of graph analytics algorithms on a modern high-end server
The scheme is open to all EEECS Undergraduates (apart from students on the BIT degree pathway and students who are due to graduate this summer)
The successful candidate should be acquainted with the C programming language, or be willing and able to learn the basics quickly. We welcome candidates on all pathways, including CS and SE students and SESE and EEE students who have an affinity to programming.
Each internship will last between 6-8 weeks and will pay a weekly stipend of £200.
Accommodation and travel costs are not provided under this scheme.
Start date: To be determined in conjunction with the successful candidate
Duration: 6-8 (Weeks) (project can be adjusted to 6 or 8 weeks depending on student availability)
Location: Computer Science Building
Further information available at: http://www.qub.ac.uk/schools/eeecs/Research/
|Supervisor Name:||Hans Vandierendonck|
Queens University of Belfast
|Tel:||+44 (0)28 9097 4654|
For further information on Research Area click on link below: