A new schedule algorithm for LTE signal processing

Scheduling applications on multi-processor is an effective core technology in using distributed systems. Task scheduling is to map applications to parallel system in order to minimize the makespan. Task scheduling is classified into static scheduling and dynamic scheduling. Most task scheduling is NP-complete problem. The authors come up with a new scheduling algorithm: Chunking Splicing Algorithm for a special type of DAG like the processing of PDSCH (Physical Downlink Shared Channel) in LTE. When mapping that special type of DAG (like signal processing of PDSCH), its performance is quite good comparing with traditional list scheduling algorithms: HLFET, ISH, ETF, DLS algorithm. Also the authors use some random DAGs to test the performance of our new algorithm.

I. Introduction

Parallel computing emerged in the early 1960s, it refers to the partition of the applications into multiple task sets which have no dependencies with each other (also called as: parallel programs), and then allocating the tasks to different processors. With the increasing number of users, the variety of business types and the expansion of business scale, the scale of the system, the processing time and the computation complexity are also increasing rapidly, the authors address the need for high performance systems by parallel computing. Thus, it's still a technique that requires continuous improvement to obtain high performance when scheduling tasks on different processors.