Use this URL to cite or link to this record in EThOS:
Title: Acceleration of algorithms for dynamic network optimisation
Author: Ye, Yuanzhou
Awarding Body: University of Reading
Current Institution: University of Reading
Date of Award: 2013
Availability of Full Text:
Access from EThOS:
With the unprecedentedly growing demand for better-faster-cheaper mobile services anytime and anywhere, radio network optimisation has witnessed a tremendous increase in its complexity and cost. The Dynamic Network Optimisation (DNO) concept emerged as a sound solution to optimally and continuously adjusting the network configurations in response to changing subscriber needs and network conditions. Yet, the realisation of DNO is still in its infancy, largely due to a lack of breakthrough in minimising the lengthy optimisation runtime. One solution is faster theoretical methods and another is the acceleration of existing optimisation algorithms. This thesis addresses the latter with the design and implementation of a set of novel parallel solutions, namely multi-threaded, high-level, distributed and cloud parallel solutions that significantly accelerate the sophisticated network optimisation algorithms whilst retaining the optimisation quality and precision. The GEOson (originally known as ariesoACP) platform serves as the testbed for parallelisation. All parallel solutions are implemented upon multi-core architectures. For the multi-threaded solution, real-life project results exhibit a promising run-time reduction between 11% and 42% and it features identical optimisation outputs compared to the sequential version. For the high-level and distributed solutions, a scalable and best possible speedup of 2.8 with 4 cores, 6 with a (imperfect-tree pattern based) distributed 10-core system and 7.5 with a distributed 16-core system, has been achieved alongside self- consistent optimisation outcome. Moreover, the acceleration is irrespective of the network size and complexity. The proposed parallel systems are designed such that they can scale up to the cloud and high-performance computing platforms seamlessly. A holistic DNO design is subsequently presented with a case study and potential solutions to the real-time DNO. Overall, the thesis has made a major breakthrough towards the realisation of DNO. Furthermore, most of the proposed methodologies and solutions exhibit applicability to various other algorithms and applications.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available