A Game-Theoretic Approach for Non-Cooperative Load Balancing Among Competing Cloudlets

Abstract

To deliver high performance and reliability to the mobile users in accessing mobile cloud services, the major interest is currently given to the integration of centralized cloud computing and distributed edge computing infrastructures. In such a heterogeneous network ecosystem, multiple cloudlets from different service providers coexist. However, to meet the stringent latency requirements of computation-intensive and mission-critical applications, overloaded cloudlets can offload some of the incoming job requests to their relatively under-loaded neighboring cloudlets. In this paper, we propose a novel economic and non-cooperative game-theoretic model for load balancing among competitive cloudlets. This model aims to maximize the utilities of all the competing cloudlets while meeting the end-to-end latency of the users. We characterize the problem as a generalized Nash equilibrium problem and investigate the existence and uniqueness of a pure-strategy Nash equilibrium. We design a variational inequality based algorithm to compute the pure-strategy Nash equilibrium. We show that all the competing cloudlets are able to maximize their utilities by employing our proposed Nash equilibrium computation offload strategy in both under-and overloaded conditions. We also show through numerical evaluations that our load balancing model outperforms some of the existing game-theoretic load balancing frameworks, especially in a highly overloaded condition.INDEX TERMS Cloudlet computing, non-cooperative load balancing, generalized Nash equilibrium, variational inequality.

Enhancing Mobile Edge Computing with Efficient Load Balancing Using Load Estimation in Ultra-Dense Network

“…Although load balancing is performed on the MEC servers, they only focus on short-term load balancing, which will lead to a ping-pong effect and cannot achieve long-term load balancing. References [ 7 , 10 ] introduced the idea of prediction in load balancing. The specific method is such that the size and processing time of UE application tasks follow an exponential distribution.…”

“…Even though the combination of UDN and MEC can bring great advantages and benefits, there are inevitably some problems to be solved, such as load balancing between MEC servers, which is mainly caused by the highly dynamic distribution of UE temporally and spatially and the randomness of UE application tasks [ 7 ]. Excessive offloading tasks to the same MEC server will lead to server overload and congestion, which will not only impact system performance and server life, but also lead to a sharp decline in term of quality of service (QoS) and quality of experience (QoE) [ 8 , 9 , 10 ].…”

“…However, it relies heavily on historical data. In addition, due to the highly dynamic MEC environment, the correlation between training data and real-time data is low [ 7 ]. Therefore, this paper innovatively proposes a fast and effective load balancing algorithm to solve the above problems.…”

AI-Enhanced Load Balancing in Federated Edge Computing Networks: Challenges and Solutions

“…This model is very efficient under moderate load conditions, but under very high load conditions it performs poorly because some of the cloudlets may violate the latency constraints and the Nash equilibrium (NE) solution becomes infeasible. Moreover, to the best of our knowledge, most of the existing frameworks were not designed for federated EC networks, which we addressed in [11], [12]. In this game, an overloaded cloudlet paid incentives for offloading job requests to a cloudlet only from a different SP.…”

“…3: Variation of utility function of EC node-i against ϕ ji . 0, ∀i, j = i ∈ C and a necessary condition Ω 2 ≥ Ω 3 max{t ui } + 1 max{µ ii } + max{t i j } − D Q , ∀i, j = i ∈ C is satisfied such that the feasiblilty of pure strategy Nash equilibrium (NE) solution is ensured [11]. Now, we can calculate each of the terms present in utility as follows:…”