Improving Scalability of Task Allocation and Scheduling in Large Distributed Real-Time Systems Using Shared Buffers

  • Authors:
  • Sharath Kodase;Shige Wang;Zonghua Gu;Kang G. Shin

  • Affiliations:
  • -;-;-;-

  • Venue:
  • RTAS '03 Proceedings of the The 9th IEEE Real-Time and Embedded Technology and Applications Symposium
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

Scheduling precedence-constrained tasks in a distributedreal-time system is an NP-hard problem. As a result,the task allocation and scheduling algorithms that usethese heuristics do not scale when applied to large distributedsystems. In this paper, we propose a novel approachthat eliminates inter-task dependencies using sharedbuffers between dependent tasks. The system correctness,with respect to data-dependency, is ensured by having eachdependent task poll the shared buffers at a fixed rate. Taskscan, therefore, be allocated and scheduled independentlyof their predecessors. To meet the timing constraints ofthe original dependent-task system, we have developed amethod to iteratively derive the polling rates based on end-to-end deadline constraints. The overheads associated withthe shared buffers and the polling mechanism are minimizedby clustering tasks according to their communication andtiming constraints. Our simulation results with the task allocationbased on a simple first-fit bin packing algorithmshowed that the proposed approach scales almost linearlywith the system size, and clustering tasks greatly reducesthe polling overhead.