Weighted Locality Sensitive Scheduling for Mitigating Noise on Multicore Clusters

Recent studies have shown that noise can be a significant problem as we scale to a large number of processors. One solution for mitigating noise is to turn off certain operating system services on the machine. However, this is typically infeasible because full-scale OS services may be needed for some applications. Furthermore, it is not a choice that an end user can make. Thus, we need an application-level solution. Building upon previous work which demonstrated the utility of within-node light-weight load balancing, we provide a more in-depth study of this technique, and provide insight based on experimentation for two different machines with very different noise signatures. Through careful enumeration of the search space of scheduler parameters, we allow our light-weight load balancer to be tunable as well as adaptive for a specific application running on a specific architecture. By doing this, we show how we can enable running scientific applications on very large number of processors.

Vivek Kale, Abhinav Bhatele and William D. Gropp, Weighted locality sensitive scheduling for mitigating noise on multicore clusters, In Proceedings of 18th annual IEEE International Conference on High Performance Computing (HiPC), 2011.

@inproceedings{KaleNoiseHiPC11,
  author = {Vivek Kale and Abhinav Bhatele and William D. Gropp},
  title = {Weighted locality sensitive scheduling for mitigating noise on multicore clusters},
  booktitle = {18th annual IEEE International Conference on High Performance Computing ({HiPC} 2011)},
  year = {2011},
  month = {December},
  location = {Bangalore, India}
}