POEM: Pricing Longer for Edge Computing in the Device Cloud

TL;DR

This paper introduces a long-term auction framework for edge computing in mobile device clouds, optimizing revenue and resource allocation over multiple rounds with a novel pricing approach.

Contribution

It formulates a multi-round auction mechanism for edge computing resource allocation considering device budgets, addressing the limitations of single-round auctions.

Findings

MAFL outperforms single round auctions by 55.6% in revenue.

MAFL outperforms existing double auctions by 68.6% in revenue.

The proposed mechanism effectively manages long-term benefits in edge computing auctions.

Abstract

Multiple access mobile edge computing has been proposed as a promising technology to bring computation services close to end users, by making good use of edge cloud servers. In mobile device clouds (MDC), idle end devices may act as edge servers to offer computation services for busy end devices. Most existing auction based incentive mechanisms in MDC focus on only one round auction without considering the time correlation. Moreover, although existing single round auctions can also be used for multiple times, users should trade with higher bids to get more resources in the cascading rounds of auctions, then their budgets will run out too early to participate in the next auction, leading to auction failures and the whole benefit may suffer. In this paper, we formulate the computation offloading problem as a social welfare optimization problem with given budgets of mobile devices, and consider pricing longer of mobile devices. This problem is a multiple-choice multi-dimensional 0-1 knapsack problem, which is a NP-hard problem. We propose an auction framework named MAFL for long-term benefits that runs a single round resource auction in each round. Extensive simulation results show that the proposed auction mechanism outperforms the single round by about 55.6% on the revenue on average and MAFL outperforms existing double auction by about 68.6% in terms of the revenue.