![Institutional Repository [SANDBOX]](images/dias.logo.png "Institutional Repository [SANDBOX]"){kind=logo}
Institutional Repository [SANDBOX]
Technical University of Crete
Technical University of Crete
EN
|
EL
| URI: | http://purl.tuc.gr/dl/dias/572D696B-919B-4BA7-9DE5-F1590DACF0C7 | ||
| Year | 2021 | ||
| Type of Item | Diploma Work | ||
| License |
|
||
| Bibliographic Citation | Anastasios Bokalidis, "Design and Implementation of a cloud based FPGA accelerator for phylogeny reconstruction", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2021 https://doi.org/10.26233/heallink.tuc.90599 | ||
| Appears in Collections |
One of the most interesting challenges through the 21st century, that researchershave to encounter is the rapid and continuous increase of data.Many fields of science and technology face problems in the management andprocessing of vast data. One of them is Biology. The process of phylogeneticanalysis of DNA, RNA, Protein, and other types of phylogenies, consumes alot of time which performs a non-linear increase while the volume of the datafor processing tends to increase. In addition, it is not only the time which isof concern to the scientists but also, the computing systems which are neededfor this purpose. Not even personal computers can eliminate this problem,but also high-performance computers are inadequate to face up vast data.The first ones have CPUs that can not surpass a threshold in speed up andparallelism and the second ones are used only for special studies and computations.In this project, there is a study on a phylogenetic analysis algorithm,RAxML, which is based on the maximum likelihood method. The purposeof this project is to optimize by accelerating some functions of RAxML whichconsume more than 80% of the total execution time and especially underthe processing of big data sets. So, the first step is to research the way thatRAxML behaves according to the input data and the second step is to designand construct hardware accelerators required for optimal performance.These accelerators are designed to be mapped and routed on FPGAs andalso on similar platforms of the Amazons’ cloud. Finally, there is a study andcomparison between the results coming from the initial algorithm and the results that come from the accelerators. Moreover, a theoretical model is introducedwhich shows the optimal performance of the accelerators and howit can affect the overall performance of the algorithm.
Copyright © DIAS 2013
