Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Strict Partitioning for Sporadic Rigid Gang Tasks

Binqi Sun, Tomasz Kloda, Marco Caccamo

TL;DR

This work introduces strict partitioning for sporadic rigid gang tasks on identical multiprocessors, partitioning processors into disjoint blocks and assigning tasks to a single partition to eliminate inter-partition interference. Within each partition, either uniprocessor or global gang schedulers can be used, enabling tighter schedulability tests and reduced pessimism compared to global analyses. The authors present the FFDV offline partitioning heuristic and two variants, SP-U and SP-G, with formal performance bounds for SP-U and targeted enhancements for SP-G. Extensive synthetic experiments and an Edge TPU case study show that strict partitioning generally yields higher schedulability than state-of-the-art global gang analyses, especially for low to medium task volumes, while also highlighting tradeoffs in processor utilization and scheduling overhead. Overall, strict partitioning offers practical benefits for task isolation, reduced migration, and tighter real-time guarantees in parallel real-time systems.

Abstract

The rigid gang task model is based on the idea of executing multiple threads simultaneously on a fixed number of processors to increase efficiency and performance. Although there is extensive literature on global rigid gang scheduling, partitioned approaches have several practical advantages (e.g., task isolation and reduced scheduling overheads). In this paper, we propose a new partitioned scheduling strategy for rigid gang tasks, named strict partitioning. The method creates disjoint partitions of tasks and processors to avoid inter-partition interference. Moreover, it tries to assign tasks with similar volumes (i.e., parallelisms) to the same partition so that the intra-partition interference can be reduced. Within each partition, the tasks can be scheduled using any type of scheduler, which allows the use of a less pessimistic schedulability test. Extensive synthetic experiments and a case study based on Edge TPU benchmarks show that strict partitioning achieves better schedulability performance than state-of-the-art global gang schedulability analyses for both preemptive and non-preemptive rigid gang task sets.

Strict Partitioning for Sporadic Rigid Gang Tasks

TL;DR

This work introduces strict partitioning for sporadic rigid gang tasks on identical multiprocessors, partitioning processors into disjoint blocks and assigning tasks to a single partition to eliminate inter-partition interference. Within each partition, either uniprocessor or global gang schedulers can be used, enabling tighter schedulability tests and reduced pessimism compared to global analyses. The authors present the FFDV offline partitioning heuristic and two variants, SP-U and SP-G, with formal performance bounds for SP-U and targeted enhancements for SP-G. Extensive synthetic experiments and an Edge TPU case study show that strict partitioning generally yields higher schedulability than state-of-the-art global gang analyses, especially for low to medium task volumes, while also highlighting tradeoffs in processor utilization and scheduling overhead. Overall, strict partitioning offers practical benefits for task isolation, reduced migration, and tighter real-time guarantees in parallel real-time systems.

Abstract

The rigid gang task model is based on the idea of executing multiple threads simultaneously on a fixed number of processors to increase efficiency and performance. Although there is extensive literature on global rigid gang scheduling, partitioned approaches have several practical advantages (e.g., task isolation and reduced scheduling overheads). In this paper, we propose a new partitioned scheduling strategy for rigid gang tasks, named strict partitioning. The method creates disjoint partitions of tasks and processors to avoid inter-partition interference. Moreover, it tries to assign tasks with similar volumes (i.e., parallelisms) to the same partition so that the intra-partition interference can be reduced. Within each partition, the tasks can be scheduled using any type of scheduler, which allows the use of a less pessimistic schedulability test. Extensive synthetic experiments and a case study based on Edge TPU benchmarks show that strict partitioning achieves better schedulability performance than state-of-the-art global gang schedulability analyses for both preemptive and non-preemptive rigid gang task sets.
Paper Structure (28 sections, 7 theorems, 29 equations, 10 figures, 1 table, 3 algorithms)

This paper contains 28 sections, 7 theorems, 29 equations, 10 figures, 1 table, 3 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Strict Partitioning for Rigid Gang Tasks
  4. Task and Platform Model
  5. Strict Partitioning Strategy
  6. Offline Partitioning
  7. Online Scheduling
  8. Illustrative Examples: Comparative Study
  9. Pessimisms in Global Gang Schedulability Analyses
  10. Non-parallel Execution
  11. Interference Overestimation
  12. Stationary Scheduling and Strict Partitioning
  13. Global and Partitioned for Non-preemptive Gangs
  14. Strict Partitioning Algorithms
  15. Offline Partitioning Heuristic
  16. ...and 13 more sections

Key Result

Theorem 5.1

A gang task set $\tau$ is schedulable on $M$ processors by SP-U if where the weighting function $W: [0,u_b] \rightarrow [0,\frac{8}{5}u_b]$ is defined as:

Figures (10)

  • Figure 1: Configuration time vs net execution time of DNN inferences on 16 Edge TPUs. "Inc-1,2,3,4" denote Inception-v1 inception_v1_2014, v2 inception_v2_v3_2016, v3 inception_v2_v3_2016, v4 inception_v4_inception_resnet_V2_2018, respectively; "Res-1,2,3" denote ResNet-50,101,152 resnetv1_50_101_152_2015, respectively.
  • Figure 2: New sources of pessimism in global gang scheduling.
  • Figure 3: Strict partitioning better than stationary scheduling.
  • Figure 4: Stationary scheduling better than strict partitioning.
  • Figure 5: Blocking in non-preemptive global and strict gang.
  • ...and 5 more figures

Theorems & Definitions (13)

  • Definition 3.1: Strict Partitioning
  • Example 4.1
  • Example 4.2
  • Example 4.3
  • Example 4.4
  • Example 4.5
  • Theorem 5.1
  • Theorem 5.2
  • Theorem 5.3
  • Lemma A.1
  • ...and 3 more