The Cobalt Scheduler and Resource Manager
Cobalt is a relatively lightweight batch scheduler and resource manager for clusters and large high performance computing systems. It currently supports the IBM BlueGene series of systems as well as the Cray XC40 series running CLE 6.0 as well as smaller general compute clusters. When available, Cobalt will use vendor-provided resource management tools, like ALPS or IBM's HLCS to help manage hardware. It also includes a simulator for modeling scheduler behavior with modified algorithms and functions and can be run off of prior accounting-log traces for testing new policies against production data as well as realtime-simulation modes for prototyping site policy filters and testing site-localization scripts in an offline environment.
Documentation
Installation
Simulation Configuration
System Configuration
Command Reference
System Specific Functionality
Development Notes
PBS-style Accounting Records
Papers
- E. Lusk and N. Desai and R. Bradshaw and A. Lusk and R. Butler, An Interoperability Approach to System Software, Tools and Libraries for Clusters, In International Journal of High Performance Computing Applications
- N. Desai, R. Bradshaw, A. Lusk, E. Lusk, and R. Butler.,Component-Based Cluster Systems Software Architecture: A Case Study ,In Proceedings of IEEE International Conference on Cluster Computing (CLUSTER04), 2004.
- R. Evard, N. Desai, J.P. Navarro, and D. Nurmi., Clusters as large-scale development facilities, In Proceedings of IEEE International Conference on Cluster Computing (CLUSTER02), 2002.