| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 2.1 MB | Adobe PDF |
Advisor(s)
Abstract(s)
The cloud relies on efficient algorithms to find resources for jobs by fulfilling the job's requirements and at the same time optimise an objective function. Utility is a measure of the client satisfaction that can be seen as an objective function maximised by schedulers based on the agreed service level agreement (SLA). We propose EcoVMBroker which can reduce energy consumption by using dynamic voltage frequency scaling (DVFS) and applying reductions of utility, different for classes of users and across ranges of resource allocations. Using efficient data structures and a hierarchical architecture, we created a scalable solution for the fast growing heterogeneous cloud. EcoVMBroker proved that we can delegate work in a hierarchical datacenter, make decisions based on summaries of resource usage collected from several nodes and still be efficient.
Description
Keywords
Virtual Machice scheduling Energy efficiency DVFS Partial utility
Citation
FERNANDES, Rodrigo; SIMÃO, José; VEIGA, Luís – EcoVMBroker: energy-aware scheduling for multi-layer datacenters. In 33rd Annual ACM Symposium on Applied Computing (ACM SAC). Pau, France: ACM DL, 2018. ISBN 978-1-4503-5191-1. Pp. 403-410
Publisher
Association for Computing Machinery