Markov Decision Process Based Energy-Efficient On-Line Scheduling for Slice-Parallel Video Decoders on Multicore Systems

TL;DR

This paper formulates an energy-efficient online scheduling problem for multicore video decoders as a Markov decision process, incorporating DVFS, workload, and traffic constraints, and proposes solutions evaluated on a realistic simulation platform.

Contribution

It introduces a rigorous MDP-based formulation for multicore video decoding scheduling that accounts for power, workload, and traffic constraints, with practical approximations to reduce complexity.

Findings

MDP formulation effectively models the scheduling problem.

Proposed algorithms reduce power consumption while maintaining QoS.

Simulation results demonstrate the approach's practicality and efficiency.

Abstract

We consider the problem of energy-efficient on-line scheduling for slice-parallel video decoders on multicore systems. We assume that each of the processors are Dynamic Voltage Frequency Scaling (DVFS) enabled such that they can independently trade off performance for power, while taking the video decoding workload into account. In the past, scheduling and DVFS policies in multi-core systems have been formulated heuristically due to the inherent complexity of the on-line multicore scheduling problem. The key contribution of this report is that we rigorously formulate the problem as a Markov decision process (MDP), which simultaneously takes into account the on-line scheduling and per-core DVFS capabilities; the power consumption of the processor cores and caches; and the loss tolerant and dynamic nature of the video decoder's traffic. In particular, we model the video traffic using a Direct Acyclic Graph (DAG) to capture the precedence constraints among frames in a Group of Pictures (GOP) structure, while also accounting for the fact that frames have different display/decoding deadlines and non-deterministic decoding complexities. The objective of the MDP is to minimize long-term power consumption subject to a minimum Quality of Service (QoS) constraint related to the decoder's throughput. Although MDPs notoriously suffer from the curse of dimensionality, we show that, with appropriate simplifications and approximations, the complexity of the MDP can be mitigated. We implement a slice-parallel version of H.264 on a multiprocessor ARM (MPARM) virtual platform simulator, which provides cycle-accurate and bus signal-accurate simulation for different processors. We use this platform to generate realistic video decoding traces with which we evaluate the proposed on-line scheduling algorithm in Matlab.