Document Details

Document Type : Thesis 
Document Title :
Load Balancing for Big Data Computing with Data Locality Awareness
موازنة العبء لحوسبة البيانات الكبيرة مع الوعي بمحلية البيانات
 
Subject : Faculty of Computing and Information Technology 
Document Language : Arabic 
Abstract : Load balancing of computational tasks and data plays a critical role in big data systems. Load balancing attempts to optimize the overall computation or system performance, such as throughput, solution (or response) time, and resource usage. This is achieved by an optimum allocation of data and workloads to the available resources. Data locality techniques aim to map tasks to the nodes where the relevant data resides. The two performance objectives (i.e. load balancing and data locality) are usually at odds. There is a need to find optimal strategies to maximize the performance. The aim of this thesis is to develop data locality aware load balancing techniques for big data computing and apply these to practical problems of high significance. A detailed review of the literature is presented to identify major challenges in big data research. These challenges include, among others, load balancing and data locality. A load balancing technique that is also aware of data locality has been developed and is applied to a graph-based road transportation problem. We have modelled the entire US road network data that contains approximately 24 million vertices and 58 million arcs; the specific aim of this research is to identify various points of interests (PoIs) in the regions including living places and healthcare centres. These PoIs are subsequently used to find the shortest paths among them for planning and operations purposes. The relevant algorithms that we have developed are presented in the thesis. The algorithms have been implemented using the Spark big data platform tools on the Aziz supercomputer, a Top500 supercomputer in the world according to the June and November 2015 rankings. The results are collected in terms of the shortest paths and the road networks are visualised as graphs. The performance of the load balancing and data locality awareness techniques is analysed against a varying number of nodes on the Aziz supercomputer and a good speedup has been reported. Conclusions are drawn with directions for future work. 
Supervisor : Prof. Rashid Mehmood 
Thesis Type : Master Thesis 
Publishing Year : 1438 AH
2017 AD 
Co-Supervisor : Dr. Aiiad Albeshri 
Added Date : Monday, June 5, 2017 

Researchers

Researcher Name (Arabic)Researcher Name (English)Researcher TypeDr GradeEmail
ياسر عرفات --, Yasir ArfatResearcherMaster 

Files

File NameTypeDescription
 40855.pdf pdf 

Back To Researches Page