Pregled bibliografske jedinice broj: 500438
Simulation of grid computing workload management using weighted random matching
Simulation of grid computing workload management using weighted random matching // Advances in production engineering & management, 6 (2011), 1; 5-14 (međunarodna recenzija, članak, znanstveni)
CROSBI ID: 500438 Za ispravke kontaktirajte CROSBI podršku putem web obrasca
Naslov
Simulation of grid computing workload management using weighted random matching
Autori
Mudnić, Eugen ; Vicković, Linda ; Čelar, Stipo
Izvornik
Advances in production engineering & management (1854-6250) 6
(2011), 1;
5-14
Vrsta, podvrsta i kategorija rada
Radovi u časopisima, članak, znanstveni
Ključne riječi
grid computing ; grid resource broker ; grid workload management ; discrete-event simulation
Sažetak
This paper presents the simulation results of a novel Grid computing workload management scheme based on job-pull operational mode and weighted random matching. Large scale Grid computing system as one used in the CERN-ALICE experiment has been successfully modelled and simulated using discrete-event simulation technique. The efficiency of the proposed scheme has been demonstrated by using the workload resembling the processing of high throughput computing tasks of the CERN-ALICE experiment. To generate more realistic synthetic workload the model with user- oriented workload generation has been used. The simulation results showed that the proposed scheme enables efficient intracommunity workload, giving each user weighted fair access to computing resources. Therefore the applied stochastic matching enables implementation of the computational economy for the workload management scheme with job-pull operational mode. However, the presented results are limited on large scale computing systems and workloads that consist of uniform and independent jobs.
Izvorni jezik
Engleski
Znanstvena područja
Računarstvo
POVEZANOST RADA
Ustanove:
Fakultet elektrotehnike, strojarstva i brodogradnje, Split
Citiraj ovu publikaciju:
Uključenost u ostale bibliografske baze podataka::
- INSPEC
- EBSCO