Heavy-Tailed Distributions: Their Impact on Noncontiguous Allocation and Job Scheduling in 2D Mesh NoCs

Multicore system on chip has led to the evolution of many research areas due to the advances in circuit technology design and performance issues. Mesh topology is the most common template that was proposed for the communication of the network on chip due to its scalability and simplest form. This paper investigates the effect of bounded pareto job size distribution on the performance of well-known non-contiguous allocation strategies (GABL, Random, MBS, Paging (0)) based on different communication patterns (Near Neighbor, One to All, and Random) and various scheduling strategies (First-Come-First-Served (FCFS), Out-of-Order (OO), and Window-Based) in 2D mesh NoCs. The results show that the performance of the non-contiguous allocation strategies is improved when the distribution of job size is bounded pareto. Also, the results for the two job size distributions considered (i.e. uniform distribution and bounded pareto distribution) show that GABL, MBS, and paging (0) allocation strategies outperform Random allocation strategy. Moreover, the results reveal that the scheduling strategies OO and window-based improve the system performance over FCFS scheduling and also near neighbor communication has the best performance among other communication patterns considered in this research work.