Oh-Trust: Overbooking and Hybrid Trading-empowered Resource Scheduling with Smart Reputation Update over Dynamic Edge Networks

TL;DR

Oh-Trust advances resource scheduling in dynamic edge networks by unifying overbooking-enabled futures trading with spot contracts and a smart, RL-based reputation update mechanism. The BiN_CDO algorithm negotiates risk-aware long-term contracts with overbooking, while BiN_TCD handles temporary spot agreements to cover unmet demand. SRU_ConR uses deep reinforcement learning to adaptively renew contracts based on market reputation, improving long-term stability. Across simulations with real-world data, Oh-Trust achieves balanced utilities, higher transaction success rates, and stronger market credibility compared with baselines, demonstrating practical potential for edge-resource sharing under uncertainty.

Abstract

Incentive-driven computing resource sharing is crucial for meeting the ever-growing demands of emerging mobile applications. Although conventional spot trading offers a solution, it frequently leads to excessive overhead due to the need for real-time trading related interactions. Likewise, traditional futures trading, which depends on historical data, is susceptible to risks from network dynamics. This paper explores a dynamic and uncertain edge network comprising a computing platform, e.g., an edge server, that offers computing services as resource seller, and various types of mobile users with diverse resource demands as buyers, including fixed buyers (FBs) and uncertain occasional buyers (OBs) with fluctuating needs. To facilitate efficient and timely computing services, we propose an overbooking- and hybrid trading-empowered resource scheduling mechanism with reputation update, termed Oh-Trust. Particularly, our Oh-Trust incentivizes FBs to enter futures trading by signing long-term contracts with the seller, while simultaneously attracting OBs to spot trading, enhancing resource utilization and profitability for both parties. Crucially, to adapt to market fluctuations, a smart reputation updating mechanism is integrated, allowing for the timely renewal of long-term contracts to optimize trading performance. Extensive simulations using real-world datasets demonstrate the effectiveness of Oh-Trust across multiple evaluation metrics.

Figures (5)

• Figure 1: Framework and procedure in terms of a timeline associated with our proposed the Oh-Trust in dynamic edge networks.
• Figure 2: A detailed structure of the SRU_ConR.
• Figure 3: Performance comparisons in terms of different parties upon having $\tilde{r}= 600$: (a) utility of buyers, (b) utility of seller, (c) PoTSU.
• Figure 4: Performance comparisons in terms of: (a) NI, (b) PTCT, and (c) RT under different problem sizes. Specifically, S1-S6 are set as $\{30/500\}$, $\{40/500\}$, $\{50/500\}$, $\{30/600\}$, $\{40/600\}$, and $\{50/600\}$.
• Figure 5: Performance comparisons in terms of: (a) VoRM, (b) TRLC, and (c) UoR under different problem sizes. Specifically, S1-S6 are set as $\{30/500\}$, $\{40/500\}$, $\{50/500\}$, $\{30/600\}$, $\{40/600\}$, and $\{50/600\}$.