Cloud computing has the potential to revolutionize scientific computing, much as it has transformed enterprise information technology. By providing an elastically scalable pool of computing resources that can be provisioned on demand, cloud computing can allow any researcher to perform even the largest data analyses. Yet cloud computing is no panacea in its current form. First, it requires significant technical knowledge to efficiently provision and manage cloud resources. Second, it can easily become expensive, even when resource provisioning systems automatically acquire resources and avoid holding onto them longer than intended.
The Scalable Cost-Aware Cloud Infrastructure Management and Provisioning (SCRIMP) project aims to address these challenges by developing new, more efficient cloud provisioning methods and integrating these new methods into automated cloud access tools. In so doing, the project will improve the complexity, cost, and efficiency of leveraging cloud computing platforms by an order of magnitude or more.
Our research focuses on three core areas: (1) developing new cloud profiling models that can efficiently predict application performance (execution time, accuracy, and resource usage) on many different cloud instance types; (2) exploring cloud market prediction models to support provisioning of volatile instances at low cost and with low risk; and (3) creating a user-oriented provisioning service that uses application profiles and market prediction methods to automate and optimize infrastructure provisioning decisions and cloud management based on user requirements (e.g., optimizing run time, reducing costs).
- R. Chard, K. Chard, B. Ng, K. Bubendorfer, A. Rodriguez, R. Madduri, and I. Foster, “An Automated Tool Profiling Service for the Cloud,” 2016 16th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGrid), Cartagena, 2016, pp. 223-232.
- R. Chard, K. Chard, K. Bubendorfer, L. Lacinski, R. Madduri and I. Foster, “Cost-Aware Cloud Provisioning,” e-Science (e-Science), 2015 IEEE 11th International Conference on, Munich, 2015, pp. 136-144.
- R. Wolski, J. Brevik, R. Chard, and K. Chard. Probabilistic guarantees of execution duration for Amazon spot instances. Under review. 2016