A Game Approach to Multi-Servers Load Balancing with Load-Dependent Server Availability Consideration

Liu, Chubo; Li, Kenli

doi:10.1109/tcc.2018.2790404

Cited by 90 publications

(60 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…• Generic edge clouds [15], [16], [19], [21], [29], [30] • Hierarchical clouds [13], [14], [20], [22]- [24], [26] • Fog clouds [18], [25], [27], [28] • Other [17] Decision variables…”

Section: Target Environmentmentioning

confidence: 99%

“…• Workload placement and resource allocation [14], [15], [19], [20], [22], [24], [26], [27], [29] • Workload placement [13], [16], [21], [23], [25], [28] • Other [17], [18], [30] Optimization objective…”

Section: Target Environmentmentioning

confidence: 99%

“…• Problem relaxation [13], [14], [17], [19], [20], [22], [25], [29] • Approximation [15], [18], [21], [23], [24], [26], [30] • Heuristic based on machine learning [27], [28] • Game theory [16] QoS awareness…”

Section: Target Environmentmentioning

confidence: 99%

“…• Best-effort [13], [16], [21]- [23], [28], [29] • Response time QoS [15], [18]- [20], [24], [25], [30] • Other [14], [17], [26], [27] related to network topology [15], [16], [19], [21], [29], [30]. Some other schemes consider hierarchical edge clouds, where a higher-tier cloud server plays a role of back-up for overloaded lower-tier servers [13], [14], [20], [22]- [24], [26].…”

Section: Target Environmentmentioning

confidence: 99%

“…Therefore, the majority of existing workload distribution schemes jointly control the workload placement and resource allocation [14], [15], [19], [20], [22], [24], [26], [27], [29]. Since such joint controls may worsen the problem tractability, many methods only decide the workload placement and then, accordingly, control the resource allocation by pre-defined equations or algorithms [13], [16], [21], [23], [25], [28]. Some other methods handle other decision variables related to specific problem domains.…”

Section: Target Environmentmentioning

confidence: 99%

See 4 more Smart Citations

QoS-Aware Workload Distribution in Hierarchical Edge Clouds: A Reinforcement Learning Approach

et al. 2020

View full text Add to dashboard Cite

Recently, edge computing is getting attention as a new computing paradigm that is expected to achieve short-delay and high-throughput task offloading for large scale Internet-of-Things (IoT) applications. In edge computing, workload distribution is one of the most critical issues that largely influences the delay and throughput performance of edge clouds, especially in distributed Function-as-a-Service (FaaS) over networked edge nodes. In this paper, we propose the Resource Allocation Control Engine with Reinforcement learning (RACER), which provides an efficient workload distribution strategy to reduce the task response slowdown with per-task response time Quality-of-Service (QoS). First, we present a novel problem formulation with the per-task QoS constraint derived from the well-known token bucket mechanism. Second, we employ a problem relaxation to reduce the overall computation complexity by compromising just a bit of optimality. Lastly, we take the deep reinforcement learning approach as an alternative solution to the workload distribution problem to cope with the uncertainty and dynamicity of underlying environments. Evaluation results show that RACER achieves a significant improvement in terms of per-task QoS violation ratio, average slowdown, and control efficiency, compared to AREA, a state-ofthe-art workload distribution method. INDEX TERMS deep reinforcement learning, edge computing, resource allocation, workload distribution

show abstract

Section: Target Environmentmentioning

confidence: 99%

Section: Target Environmentmentioning

confidence: 99%

Section: Target Environmentmentioning

confidence: 99%

Section: Target Environmentmentioning

confidence: 99%

Section: Target Environmentmentioning

confidence: 99%

See 3 more Smart Citations

QoS-Aware Workload Distribution in Hierarchical Edge Clouds: A Reinforcement Learning Approach

et al. 2020

View full text Add to dashboard Cite

show abstract

Multilevel controller‐assisted intrinsically modified ant colony optimization heuristic‐based load‐balancing model for mega cloud infrastructures

Govardhan

Srinivasan

2022

Int J Communication

View full text Add to dashboard Cite

Summary The exponential rise in decentralized computing demands has made cloud computing an inevitable technical need; however, under exceedingly high load dynamism, ensuring service level agreement (SLA) has remained a challenge. Despite of the fact that heuristic‐driven VM migration or load‐balancing methods are effective towards quality of service (QoS) and reliable cloud services, the major existing solutions ignore the issues of local minima and convergence and hence undergo SLA violation (SLAV). Moreover, most of the existing methods are designed as a reactive scheduling concept that often results into delayed scheduling and hence SLAV or QoS compromise. To alleviate such issues, in this paper, a robust intrinsically modified evolutionary computing‐driven dynamic load‐balancing model (intrinsically modified ant colony system [IMACS]) is proposed for mega cloud infrastructures. To address aforesaid issues, IMACS employs stochastic prediction‐based overload detection, followed by SLA‐sensitive VM selection and IMACS for load balancing. In IMACS, the use of stochastic prediction‐based overload detection method enabled timely hotspot detection and hence improved proactive decision ability. Moreover, it applies minimum migration time‐sensitive VM selection concept that too helped in guaranteeing load balancing without undergoing downtime or SLAV. To improve VM migration, unlike classical heuristic methods, we designed IMACS with multitype population, adaptive pheromone update, adaptive pheromone diffusion, and coevolution concept, which not only helped in reducing the search space and allied computation but also enabled time‐efficient scheduling decision. The overall proposed model was realized in sync with two controllers named local controller and global controller. Here, the first controller collects the real‐time VM and VM manager (VMM) information which is used to detect overloaded host and the VM to be migrated. While the global controller is designed with IMACS to exploit the information from local controller to execute QoS‐sensitive VM migration. The proposed IMACS‐driven load‐balancing model was realized with CloudSim platform were simulating it over 800 number of independent VMs, each encompassing autonomous task, the performance was analyzed in terms of SLAV, number of migrations, SLA per active host, and energy consumption. Simulation results revealed that the proposed model exhibits 20% higher migration, while retaining almost 19% better SLA in comparison to the other state‐of‐art methods like genetic algorithm or classical ant colony system‐based load‐balancing methods. IMACS exhibited 30% higher host shutdown signifying better energy efficiency (15% lower power consumption).

show abstract

A Game Approach to Multi-Servers Load Balancing with Load-Dependent Server Availability Consideration

Cited by 90 publications

References 32 publications

QoS-Aware Workload Distribution in Hierarchical Edge Clouds: A Reinforcement Learning Approach

QoS-Aware Workload Distribution in Hierarchical Edge Clouds: A Reinforcement Learning Approach

Multilevel controller‐assisted intrinsically modified ant colony optimization heuristic‐based load‐balancing model for mega cloud infrastructures

Optimal load distribution for multiple classes of applications on heterogeneous servers with variable speeds

Contact Info

Product

Resources

About