There is some work being done in CMU about scheduling for heterogeneous clusters. Tetrisched takes into job specific specifications and estimated run-times in order to maximize throughput. For example, it will run a job that is concerned with completion on a GPU, it will run two jobs that communicate with each other on the same machine etc.
There is some work being done in CMU about scheduling for heterogeneous clusters. Tetrisched takes into job specific specifications and estimated run-times in order to maximize throughput. For example, it will run a job that is concerned with completion on a GPU, it will run two jobs that communicate with each other on the same machine etc.
http://www.pdl.cmu.edu/PDL-FTP/CloudComputing/CMU-PDL-13-112.pdf