A parallel event system for large-scale cloud simulations in DISSECT-CF

Discrete Event Simulation (DES) frameworks gained significant popularity to support and evaluate cloud computing environments. They support decision-making for complex scenarios, saving time and effort. The majority of these frameworks lack parallel execution. In spite being a sequential framework,...

Teljes leírás

Elmentve itt :
Bibliográfiai részletek
Szerzők: Sallo Dilshad Hassan
Kecskeméti Gábor
Testületi szerző: Conference of PhD Students in Computer Science (12.) (2020) (Szeged)
Dokumentumtípus: Cikk
Megjelent: University of Szeged, Institute of Informatics Szeged 2021
Sorozat:Acta cybernetica 25 No. 2
Kulcsszavak:Számítástechnika
Tárgyszavak:
doi:10.14232/actacyb.287937

Online Access:http://acta.bibl.u-szeged.hu/75619
LEADER 02518nab a2200253 i 4500
001 acta75619
005 20220513081337.0
008 220513s2021 hu o 0|| eng d
022 |a 0324-721X 
024 7 |a 10.14232/actacyb.287937  |2 doi 
040 |a SZTE Egyetemi Kiadványok Repozitórium  |b hun 
041 |a eng 
100 1 |a Sallo Dilshad Hassan 
245 1 2 |a A parallel event system for large-scale cloud simulations in DISSECT-CF  |h [elektronikus dokumentum] /  |c  Sallo Dilshad Hassan 
260 |a University of Szeged, Institute of Informatics  |b Szeged  |c 2021 
300 |a 469-484 
490 0 |a Acta cybernetica  |v 25 No. 2 
520 3 |a Discrete Event Simulation (DES) frameworks gained significant popularity to support and evaluate cloud computing environments. They support decision-making for complex scenarios, saving time and effort. The majority of these frameworks lack parallel execution. In spite being a sequential framework, DISSECT-CF introduced significant performance improvements when simulating Infrastructure as a Service (IaaS) clouds. Even with these improvements over the state of the art sequential simulators, there are several scenarios (e.g., large scale Internet of Things or serverless computing systems) which DISSECT-CF would not simulate in a timely fashion. To remedy such scenarios this paper introduces parallel execution to its most abstract subsystem: the event system. The new event subsystem detects when multiple events occur at a specific time instance of the simulation and decides to execute them either on a parallel or a sequential fashion. This decision is mainly based on the number of independent events and the expected workload of a particular event. In our evaluation, we focused exclusively on time management scenarios. While we did so, we ensured the behaviour of the events should be equivalent to realistic, larger-scale simulation scenarios. This allowed us to understand the effects of parallelism on the whole framework, while we also shown the gains of the new system compared to the old sequential one. With regards to scaling, we observed it to be proportional to the number of cores in the utilised SMP host. 
650 4 |a Természettudományok 
650 4 |a Számítás- és információtudomány 
695 |a Számítástechnika 
700 0 1 |a Kecskeméti Gábor  |e aut 
710 |a Conference of PhD Students in Computer Science (12.) (2020) (Szeged) 
856 4 0 |u http://acta.bibl.u-szeged.hu/75619/1/cybernetica_025_numb_002_469-484.pdf  |z Dokumentum-elérés