Dynamic Task Fetching Over Time Varying Wireless Channels for Mobile Computing Applications

TL;DR

This paper addresses adaptive task prefetching for mobile devices over unreliable wireless channels, proposing algorithms that optimize latency and buffer usage with near-optimal performance.

Contribution

It introduces a dynamic programming model and two prefetching algorithms, FON and RFON, for efficient task fetching under varying wireless and memory conditions.

Findings

Algorithms achieve near-optimal latency performance.

Proposed methods adapt well to slow and fast fading scenarios.

Reduced computational complexity compared to full dynamic optimization.

Abstract

The processing, computation and memory requirements posed by emerging mobile broadband services require adaptive memory management and prefetching techniques at the mobile terminals for satisfactory application performance and sustained device battery lifetime. In this work we investigate a scenario where tasks with varied computational requirements are fetched by a mobile device from a central server over an error prone wireless link. We examine the buffer dynamics at the mobile terminal and the central server under varying wireless channel connectivity and device memory congestion states as variable sizes tasks are executed on the terminal. Our goal is to minimize the latency experienced by these tasks while judiciously utilizing the device buffering capability. We use a dynamic programming framework to model the optimal prefetching policy. We further propose a) a prefetching algorithm Fetch-or- Not (FON), which uses quasi-static assumption on system state to make prefetching decisions, and b) a prefetching policy RFON, which uses randomized approximation to the optimal solution thus obviating the need for dynamic online optimization and substantially reducing the computational complexity. Through performance evaluation under slow and fast fading scenarios we show that proposed algorithms come close to performance of the optimal scheme.