Optimal provisioning for scheduling divisible loads with reserved cloud resources
Cloud computing offers customers an efficient way to flexibly allocate resources to meet demands. Cloud service vendors can offer consumers three purchasing plans, i.e., on-demand, spot, and reserved instances for resource provisioning. Since price of resources in reservation plan is generally cheap...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Conference Paper |
| Language: | English |
| Published: |
2013
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/102033 http://hdl.handle.net/10220/16390 |
| _version_ | 1826113329392779264 |
|---|---|
| author | Hu, Menglan. Luo, Jun. Veeravalli, Bharadwaj. |
| author2 | School of Computer Engineering |
| author_facet | School of Computer Engineering Hu, Menglan. Luo, Jun. Veeravalli, Bharadwaj. |
| author_sort | Hu, Menglan. |
| collection | NTU |
| description | Cloud computing offers customers an efficient way to flexibly allocate resources to meet demands. Cloud service vendors can offer consumers three purchasing plans, i.e., on-demand, spot, and reserved instances for resource provisioning. Since price of resources in reservation plan is generally cheaper than that in on-demand plan, in this study we attempt to make use of the cheap reserved instances to reduce monetary costs. We consider processing a large divisible load onto on-demand and reserved instances in clouds. Divisible loads, also called embarrassingly parallel workloads, can be partitioned into an arbitrarily large number of independent load fractions and be distributed across multiple processing nodes. We investigate the time-cost optimization problems for provisioning resources and scheduling divisible loads with reserved instances in clouds. The objectives are two-fold: First, given a total processing time (deadline), minimize the total cost. Second, given a budget (total cost), minimize the total processing time. We formulate the problems as mixed integer programs (MIP). We show that the optimal solutions of the problems have very simple structures. We then propose light-weight optimal solutions for the problems with rigorous proofs. Numerical experiments are presented to illustrate the salient features of these solutions. |
| first_indexed | 2024-10-01T03:21:25Z |
| format | Conference Paper |
| id | ntu-10356/102033 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T03:21:25Z |
| publishDate | 2013 |
| record_format | dspace |
| spelling | ntu-10356/1020332020-05-28T07:18:25Z Optimal provisioning for scheduling divisible loads with reserved cloud resources Hu, Menglan. Luo, Jun. Veeravalli, Bharadwaj. School of Computer Engineering IEEE International Conference on Networks (18th : 2012 : Singapore) DRNTU::Engineering::Computer science and engineering Cloud computing offers customers an efficient way to flexibly allocate resources to meet demands. Cloud service vendors can offer consumers three purchasing plans, i.e., on-demand, spot, and reserved instances for resource provisioning. Since price of resources in reservation plan is generally cheaper than that in on-demand plan, in this study we attempt to make use of the cheap reserved instances to reduce monetary costs. We consider processing a large divisible load onto on-demand and reserved instances in clouds. Divisible loads, also called embarrassingly parallel workloads, can be partitioned into an arbitrarily large number of independent load fractions and be distributed across multiple processing nodes. We investigate the time-cost optimization problems for provisioning resources and scheduling divisible loads with reserved instances in clouds. The objectives are two-fold: First, given a total processing time (deadline), minimize the total cost. Second, given a budget (total cost), minimize the total processing time. We formulate the problems as mixed integer programs (MIP). We show that the optimal solutions of the problems have very simple structures. We then propose light-weight optimal solutions for the problems with rigorous proofs. Numerical experiments are presented to illustrate the salient features of these solutions. 2013-10-10T06:07:26Z 2019-12-06T20:48:34Z 2013-10-10T06:07:26Z 2019-12-06T20:48:34Z 2012 2012 Conference Paper Hu, M., Luo, J., & Veeravalli, B. (2012). Optimal provisioning for scheduling divisible loads with reserved cloud resources. 2012 18th IEEE International Conference on Networks (ICON), 204-209. https://hdl.handle.net/10356/102033 http://hdl.handle.net/10220/16390 10.1109/ICON.2012.6506559 en |
| spellingShingle | DRNTU::Engineering::Computer science and engineering Hu, Menglan. Luo, Jun. Veeravalli, Bharadwaj. Optimal provisioning for scheduling divisible loads with reserved cloud resources |
| title | Optimal provisioning for scheduling divisible loads with reserved cloud resources |
| title_full | Optimal provisioning for scheduling divisible loads with reserved cloud resources |
| title_fullStr | Optimal provisioning for scheduling divisible loads with reserved cloud resources |
| title_full_unstemmed | Optimal provisioning for scheduling divisible loads with reserved cloud resources |
| title_short | Optimal provisioning for scheduling divisible loads with reserved cloud resources |
| title_sort | optimal provisioning for scheduling divisible loads with reserved cloud resources |
| topic | DRNTU::Engineering::Computer science and engineering |
| url | https://hdl.handle.net/10356/102033 http://hdl.handle.net/10220/16390 |
| work_keys_str_mv | AT humenglan optimalprovisioningforschedulingdivisibleloadswithreservedcloudresources AT luojun optimalprovisioningforschedulingdivisibleloadswithreservedcloudresources AT veeravallibharadwaj optimalprovisioningforschedulingdivisibleloadswithreservedcloudresources |