Virtualization of 5G Cellular Networks as a Hierarchical Combinatorial Auction (1511.08256v1)

Abstract: Virtualization has been seen as one of the main evolution trends in the forthcoming fifth generation (5G) cellular networks which enables the decoupling of infrastructure from the services it provides. In this case, the roles of infrastructure providers (InPs) and mobile virtual network operators (MVNOs) can be logically separated and the resources (e.g., subchannels, power, and antennas) of a base station owned by an InP can be transparently shared by multiple MVNOs, while each MVNO virtually owns the entire BS. Naturally, the issue of resource allocation arises. In particular, the InP is required to abstract the physical resources into isolated slices for each MVNO who then allocates the resources within the slice to its subscribed users. In this paper, we aim to address this two-level hierarchical resource allocation problem while satisfying the requirements of efficient resource allocation, strict inter-slice isolation, and the ability of intra-slice customization. To this end, we design a hierarchical combinatorial auction mechanism, based on which a truthful and sub-efficient resource allocation framework is provided. Specifically, winner determination problems (WDPs) are formulated for the InP and MVNOs, and computationally tractable algorithms are proposed to solve these WDPs. Also, pricing schemes are designed to ensure incentive compatibility. The designed mechanism can achieve social efficiency in each level even if each party involved acts selfishly. Numerical results show the effectiveness of the proposed scheme.

• The paper introduces a hierarchical auction framework that allocates 5G network resources by decoupling physical infrastructure from virtual service provisions.
• It employs dynamic programming and greedy algorithms to solve NP-hard winner determination problems while ensuring incentive compatibility.
• Numerical results demonstrate that the approach outperforms fixed and general sharing benchmarks, enhancing both network efficiency and user satisfaction.

Overview of Hierarchical Combinatorial Auction Mechanism for 5G Network Virtualization

The paper, authored by Kun Zhu and Ekram Hossain, presents a sophisticated framework for resource allocation in 5G cellular networks through virtualization. This virtualization is accomplished by treating the network infrastructure and service provision separately, thus allowing base station resources to be shared among multiple Mobile Virtual Network Operators (MVNOs) who can virtualize the ownership of base stations (BS). The complexity of efficiently allocating resources—such as subchannels, power, and antennas—necessitates an innovative hierarchical combinatorial auction mechanism.

Hierarchical Resource Allocation

The framework addresses two primary hierarchical levels within the resource allocation paradigm: the Infrastructure Provider (InP) level and the MVNO level. Each level is integrated into a broader hierarchical combinatorial auction model. At the InP level, the physical resources owned by InPs are allocated in bundles or slices to MVNOs. Subsequently, MVNOs manage these resources by further slicing them for their own users. The hierarchical model underpins strict inter-slice isolation while allowing flexibility for intra-slice customization through demand-driven slicing.

Mechanism Design and Core Components

This paper's novelty lies in its auction-based approach that not only ensures efficient resource distribution but also maintains social efficiency, even with self-interested agents. The authors employ winner determination problems (WDPs) and design pricing schemes ensuring incentive compatibility—crucial for the auction's integrity and efficiency.

1. Winner Determination Problem (WDP): Aimed at maximizing social welfare by determining winning bids. Each winner determination problem is formulated as an integer programming optimization task known to be NP-hard. To combat high computation complexity, dynamic programming and greedy algorithms are proposed for approximate solutions.
2. Pricing Strategy: The combination of Vickrey-Clarke-Groves (VCG) principle with a base access price uniquely positions the framework to balance incentive compatibility and revenue optimization. This strategy incentivizes truthful bidding and ensures resource allocation aligns with agents' actual valuations.

Numerical Findings and Implications

The framework demonstrates superior performance compared to fixed sharing and general sharing benchmarks, as evidenced by numerical results focusing on social welfare, resource utilization, and user satisfaction. This suggests significant potential for wireless virtualization, where dynamic resource allocation dramatically enhances efficiency.

Furthermore, extensions to accommodate multiple sellers and buyers were explored, allowing for granularity in real-world applications where users might opt for varied MVNOs and MVNOs choosing diverse InPs. The multiple-seller model introduces yet another layer of flexibility, impacting user association dynamics and overall network efficiency.

Future Directions

This paper sets the stage for further research in optimizing auction-based resource allocation mechanisms. Future studies might focus on refining pricing strategies to simultaneously maximize social welfare and seller revenues more effectively. Additional examinations could also consider agent-specific valuation functions integrating fairness metrics to tackle resource allocation disputes pragmatically.

In conclusion, the proposed hierarchical combinatorial auction framework represents a substantive contribution to 5G network virtualization, promising enhanced service delivery and resource efficiency. As mobile technologies continue to evolve, such mechanisms will become critical in managing increasingly complex networks and diversified service demands.