On-line Scheduling Heuristics in Distributed Environments
Autor: | Poluta, Vlaho |
---|---|
Přispěvatelé: | Jakobović, Domagoj |
Jazyk: | angličtina |
Rok vydání: | 2016 |
Předmět: |
Raspoređivanje
paralelni strojevi heuristike raspoređivanja izvođenje u realnom vremenu raspodijeljena okolina ograničenja u resursima ograničenja pridruživanja poslova tehnike strojnog učenja genetsko programiranje suradnička koevolucija umjetne neuronske mreže machine learning techniques Scheduling machine eligibility restrictions TEHNIČKE ZNANOSTI. Računarstvo on-line execution distributed environment resource constrained scheduling heuristics TECHNICAL SCIENCES. Computing genetic programming artificial neural networks multiple machines |
Popis: | Rad se bavi specifičnim tipom raspoređivanja na paralelnim strojevima. Bavi se raspoređivanjem poslova na izvršne čvorove kroz mrežu servera za raspoređivanje. Pri tome je cilj optimizacija vremena trajanja. U ovom problemu izvršni čvorovi predstavljaju strojeve sa ograničenjima pridruživanja poslova. Svaki posao je ograničen na samo jedan stroj, a početak izvođenja mu može ovisiti o nekom drugom zadatku. Rad opisuje tri faze rješavanja problema. U svakoj od faza je predstavljen dio sustava i neke rukom pisane heuristike koje su korištene u rješavanju problema. Rad isto predstavlja neke tehnike strojnog učenja poput genetskog programiranja i neuronskih mreža koje su korištene da bi se proizvele što bolje heuristike. This thesis tackles a specific type of multiple machine scheduling problem. It deals with scheduling tasks on executing nodes through a network of scheduling servers, where the goal is to optimize the makespan. In this problem executing nodes represent the machines with eligibility restrictions. Tasks are machine bound because every task can only be executed at a specific machine and they can also be precedence constrained to other tasks. The thesis describes the tackling of the problem in three phases. Each of those phases presents a part of the system and some hand written heuristics that were used in order to solve the problem. The thesis also presents some machine learning techniques, like genetic programming and neural networks, which were used in order to produce the best possible heuristics. |
Databáze: | OpenAIRE |
Externí odkaz: |