“Data exchange” service offers individual users metadata transfer in several different formats. Citation formats are offered for transfers in texts as for the transfer into internet pages. Citation formats include permanent links that guarantee access to cited sources. For use are commonly structured metadata schemes : Dublin Core xml and ETUB-MS xml, local adaptation of international ETD-MS scheme intended for use in academic documents.
Export
Duli, Sidita
Ап approach for the parallelization of the Weibull distribution parameter estimators with applications
Autorstvo-Nekomercijalno-Bez prerade 3.0 Srbija (CC BY-NC-ND 3.0)
Academic metadata
Phd. theses
Tehnicko-tehnološke nauke
doktor nauka - elektrotehnika i računarstvo
Univerzitet Crne Gore
Elektrotehnički fakultet
Studijski program Elektronika
Other Theses Metadata
[S. Duli]
PDF/A (X, 125 pages)
Krstajić, Božo, 1968- (mentor)
Radonjić, Milutin, 1966- (član komisije)
Gajin, Slavko (član komisije)
Kaljaj, David, 1971- (član komisije)
Đukanović, Slobodan, 1976- (član komisije)
A distributed System is a collection of independent computers that appears to its users as a single coherent system. This definition has several important aspects. The first one is that it consists of autonomous computers. A second aspect is that users think they are dealing with a single system. This means that, one way or the other; the autonomous components need to collaborate. How to establish this collaboration lies at the heart of developing distributed systems.
One of the benefits of parallel computing is processing a statistical problem in less time, thus leading to a faster result. Many multidisciplinary scientific fields, such as bioinformatics, biochemistry, electrical engineering and physics, use parallel computing and distributed computing resources for simulations of experiments. The primary focus of many researches in the area of distributed Computer scheduling is finding a way to distribute tasks among the CPU cores in order to achieve better performance, such as minimizing job execution time, minimizing communication and maximizing resource utilization. Complex tasks are numerical modeling or numerical weather prediction and the specific work is to analyze the data collected from the weather stations. The most important and intensive task is the wind analysis, with its two parameters, the speed and the direction.
Different tests should be performed to determine the Weibull parameters for a particular location. In this research, the specific locations are two cities, Podgorica and Shkoder. A parallel version for estimating the Weibull parameters could be necessary to improve the quality of performing the calculations by processing them faster. It is important to compare and analyze different methods of parallelization, in order to find the adequate one which executes faster the specific algorithm.
A parallel version is built in C by using the MPI library, which implements the parallelization using the message passing architecture. Another implementation of parallel version is by using the concept of threads implemented in shared memory architecture. A third implementation in parallel of estimation of Weibull parameters uses the OpenMP directives to synchronize tasks split to different threads. A forth version is the hybrid MPI/OpenMP version. The main reason this version is implemented as the last one, is the need to compare a hybrid mode with the pure MPI implementation.
The main aim of this thesis is to analyze and compare the methods that enable the parallelization of the Weibull parameter estimation. This analysis leads to an improved version of the Weibull parameter estimation, by using the most efficient method among those compared. Also, these methods are compared with the serial version of the Weibull distribution parameter estimator, by analyzing the speed-up and the time spent in calculations.
621.39:004(043.3)
English
8374285
Tekst.
A distributed System is a collection of independent computers that appears to its users as a single coherent system. This definition has several important aspects. The first one is that it consists of autonomous computers. A second aspect is that users think they are dealing with a single system. This means that, one way or the other; the autonomous components need to collaborate. How to establish this collaboration lies at the heart of developing distributed systems.
One of the benefits of parallel computing is processing a statistical problem in less time, thus leading to a faster result. Many multidisciplinary scientific fields, such as bioinformatics, biochemistry, electrical engineering and physics, use parallel computing and distributed computing resources for simulations of experiments. The primary focus of many researches in the area of distributed Computer scheduling is finding a way to distribute tasks among the CPU cores in order to achieve better performance, such as minimizing job execution time, minimizing communication and maximizing resource utilization. Complex tasks are numerical modeling or numerical weather prediction and the specific work is to analyze the data collected from the weather stations. The most important and intensive task is the wind analysis, with its two parameters, the speed and the direction.
Different tests should be performed to determine the Weibull parameters for a particular location. In this research, the specific locations are two cities, Podgorica and Shkoder. A parallel version for estimating the Weibull parameters could be necessary to improve the quality of performing the calculations by processing them faster. It is important to compare and analyze different methods of parallelization, in order to find the adequate one which executes faster the specific algorithm.
A parallel version is built in C by using the MPI library, which implements the parallelization using the message passing architecture. Another implementation of parallel version is by using the concept of threads implemented in shared memory architecture. A third implementation in parallel of estimation of Weibull parameters uses the OpenMP directives to synchronize tasks split to different threads. A forth version is the hybrid MPI/OpenMP version. The main reason this version is implemented as the last one, is the need to compare a hybrid mode with the pure MPI implementation.
The main aim of this thesis is to analyze and compare the methods that enable the parallelization of the Weibull parameter estimation. This analysis leads to an improved version of the Weibull parameter estimation, by using the most efficient method among those compared. Also, these methods are compared with the serial version of the Weibull distribution parameter estimator, by analyzing the speed-up and the time spent in calculations.