Unifying Futures and Spot Market: Overbooking-Enabled Resource Trading in Mobile Edge Networks

TL;DR

This paper proposes a hybrid futures and spot market with overbooking for resource trading in mobile edge networks, improving resource utilization, reducing latency, and increasing profits through a novel negotiation scheme.

Contribution

It introduces an overbooking-enabled hybrid market model unifying futures and spot trading, with optimized contracts and negotiation schemes for mobile edge resource management.

Findings

Achieves higher resource utilization and profit margins.

Reduces decision-making latency and energy consumption.

Outperforms baseline methods in key performance metrics.

Abstract

Securing necessary resources for edge computing processes via effective resource trading becomes a critical technique in supporting computation-intensive mobile applications. Conventional onsite spot trading could facilitate this paradigm with proper incentives, which, however, incurs excessive decision-making latency/energy consumption, and further leads to underutilization of dynamic resources. Motivated by this, a hybrid market unifying futures and spot is proposed to facilitate resource trading among an edge server (seller) and multiple smart devices (buyers) by encouraging some buyers to sign a forward contract with seller in advance, while leaving the remaining buyers to compete for available resources with spot trading. Specifically, overbooking is adopted to achieve substantial utilization and profit advantages owing to dynamic resource demands. By integrating overbooking into futures market, mutually beneficial and risk-tolerable forward contracts with appropriate overbooking rate can be achieved relying on analyzing historical statistics associated with future resource demand and communication quality, which are determined by an alternative optimization-based negotiation scheme. Besides, spot trading problem is studied via considering uniform/differential pricing rules, for which two bilateral negotiation schemes are proposed by addressing both non-convex optimization and knapsack problems. Experimental results demonstrate that the proposed mechanism achieves mutually beneficial players' utilities, while outperforming baseline methods on critical indicators, e.g., decision-making latency, resource usage, etc.