An Efficient Dynamic Scheduling Algorithm for Multiprocessor Real-Time Systems
IEEE Transactions on Parallel and Distributed Systems
MapReduce: simplified data processing on large clusters
OSDI'04 Proceedings of the 6th conference on Symposium on Opearting Systems Design & Implementation - Volume 6
Just-In-Time Scheduling for Loop-based Speculative Parallelization
PDP '08 Proceedings of the 16th Euromicro Conference on Parallel, Distributed and Network-Based Processing (PDP 2008)
A Dynamic MapReduce Scheduler for Heterogeneous Workloads
GCC '09 Proceedings of the 2009 Eighth International Conference on Grid and Cooperative Computing
Prediction-based auto-scaling of scientific workflows
Proceedings of the 9th International Workshop on Middleware for Grids, Clouds and e-Science
| Hi-index | 0.01 |
Scheduling for speculative parallelization is a problem that remained unsolved despite its importance [2]. In the previous work scheduling was done based on Fixed-Size Chunking (FSC) technique which needed several 'dry-runs' before an acceptable finalized chunk size that will be scheduled to each processors is found. There are many other scheduling methods which were originally designed for loops with no dependences, but they were primarily focused in the problem of load balancing. In this work we address the problem of scheduling tasks with and without dependences for speculative execution. We have found that a complexity between minimizing the number of re-executions and reducing overheads can be found if the size of the scheduled block of iterations is calculated at runtime. We introduce here a scheduling method called Parallelization of Resource scheduling (PRS) in which we first analyze the processing speed of each worker based on that further division of the actual task will be done. The result shows a 5% to 10% speedup improvement in real applications with dependences with respect to a carefully tuned PRS strategy.