Thursday 30
Cloud and cluster computing
B. Dorronsoro
› 11:10 - 11:30 (20min)
A Two-level Job Scheduler for Large Distributed Systems
Bernabe Dorronsoro  1, *@  , Sergio Nesmachnow  2@  
1 : University of Luxembourg  (UL)  -  Website
2 : University of the Republic  (UdelaR)  -  Website
* : Corresponding author

Datacenters are typically composed by a huge number of computational resources, including high performance clusters, large storage systems, and/or components of large grids or cloud systems [1]. This extended abstract presents two novel two-level algorithms for scheduling a large number of workloads in a heterogeneous distributed system composed of multi-core processors. In the higher- level (i.e., between distributed systems), our schedulers implement either the traditional Round Robin (RR) and Load Balancing (LB) techniques to distribute the workflows. In the lower-level (i.e., within each datacenter), we propose a novel scheduler technique, based on HEFT and using backfilling, to schedule the locally assigned workflow using the available local resources. 

Online user: 1 RSS Feed