A Scalable Approach to SDN Control Plane Management: High Utilization Comes With Low Latency

Huang, Victoria; Chen, Gang; Zhang, Peng; Li, Hao; Hu, Chengchen; Pan, Tian; Fu, Qiang

doi:10.1109/tnsm.2020.2973222

Cited by 25 publications

(18 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recent studies have shown that CPP cannot be solved in isolation and any effective CPP solution may heavily depend on the solutions to other problems such as resource or request scheduling, load balancing, security, etc. This was achieved in [43], while in [51], Huang et.al. proposed a framework for controller placement and scheduling problem (CPSP) that addresses the two problems collectively.…”

Section: ) Genetic Algorithm Based Cppmentioning

confidence: 97%

“…The approach is a constrained optimization problem that enhances the control plane usage while ensuring low network response time. The controller scheduling problem (CPS) was solved effectively using a gradient-descent-based (GD) scheduling algorithm to optimize request distribution probability over all controllers to balance the scheduling performance and problem scalability trade-off [51]. In the same vein, the CPP was addressed using a Clustering-based GA with Cooperative Clusters (CGA-CC).…”

Section: ) Genetic Algorithm Based Cppmentioning

confidence: 99%

“…Thus, efficient tools that support real-world testbed need to be designed and developed for SDN-based sensor nodes. In the same vein, the performance of the GDbased scheduling algorithm and CGA-CC algorithm has to be evaluated on a real-world testbed to validate the effectiveness of the approach in [51]. This is important to demonstrate that CPP cannot be achieved in isolation but collectively since it may depend heavily on other related problems.…”

Section: ) Cost Analysis and Real-world Testbedsmentioning

confidence: 99%

See 2 more Smart Citations

Comprehensive Review of SDN Controller Placement Strategies

et al. 2020

View full text Add to dashboard Cite

Software-Defined Networking (SDN) is a network paradigm introduced to overcome the inherent challenges of traditional networks. Its architecture is either deployed with a single controller or multiple controllers. While the first is not suitable for large-scale networks, the latter is confronted with a controller placement problem (CPP) in a large-scale network environment. CPP involves the challenge of deploying the optimal number of controllers within a network while meeting certain performance requirements considered conflicting in nature such as reliability, load balancing, latency, energy efficiency, and computation time. A single optimal or random placement may not be feasible in CPP and careful planning is of the essence to find an appropriate trade-off among the metrics. To achieve this, several CPP approaches have been proposed, developed, and deployed over the years, each having its unique objectives, strengths, and weaknesses. Therefore, this paper performed a comprehensive review of some of the existing approaches to identify the unique solutions offered, comprehend the different strategies and the challenges that exist as well as provide researchers with future directions aimed at improving the optimum location and allocation of controllers, in particular, for SDN application in wireless sensor network (WSN). The findings revealed several existing solutions and algorithms as well as several challenges such as the need for an efficient algorithm, attack-aware, cost-aware, and energy-aware CPP schemes while achieving a good quality of service.

show abstract

Section: ) Genetic Algorithm Based Cppmentioning

confidence: 97%

Section: ) Genetic Algorithm Based Cppmentioning

confidence: 99%

Section: ) Cost Analysis and Real-world Testbedsmentioning

confidence: 99%

See 1 more Smart Citation

Comprehensive Review of SDN Controller Placement Strategies

et al. 2020

View full text Add to dashboard Cite

show abstract

“…Huang et al [105] introduce a clustering based GA approach to the CPP to partition the network into different domains and place controllers in each domain. They also propose a gradient descent scheduling algorithm to balance loads among different domains in the network.…”

Section: Ga-based Heuristic Approaches To the Cppmentioning

confidence: 99%

“…A GA [47,48] is a meta-heuristic algorithm inspired by the natural selection theory to the optimization problems. Several research efforts have been conducted on the GA based approaches to the CPP with the aim of different objectives and constraints to show the benefits of the GA in solving the CPP reviewed in Section 2.5 [94][95][96]98,99,[103][104][105]. Based on those references, GA based approaches are beneficial to find solutions that improve network performance in terms of delay among controllers and network devices and load balancing among controllers.…”

Section: Introductionmentioning

confidence: 99%

Cost-Effective Algorithms for Controller Placement Problem in Software-Defined Multihop Wireless Networks

Zahmatkesh¹

View full text Add to dashboard Cite

One key feature of software-defined networking (SDN) is a centralized control plane. However, having a central controller has some disadvantages, e.g., scalability and reliability, especially for multihop wireless networks (MWNs). Different models with multiple controllers have been proposed, each controller managing a smaller domain and collaborating with others. In an SDN architecture, solving the controller placement problem (CPP) in a multi-controller environment plays an important role on network performance in terms of delay, reliability, control overhead, etc. In this thesis, control overhead, referred to as the network cost, consists of controller-device communications to discover the network topology, exchange configurations and set up flow tables as well as inter-controller communications, if needed, to synchronize different network views and achieve the global view of the network. In software-defined multihop wireless networking (SDMWN), because of the capacity limitation and the effect of interference on wireless links, and an in-band architecture in some types of networks to exchange both data and control traffic, it is important to solve the CPP while minimizing the control overhead to reduce energy consumption, have lower packet losses and improve reliability. Therefore, the objective of the thesis is to solve the CPP in SDMWN while minimizing the number of required control packets to be exchanged in the control plane. The novelty of this work is proposing two different nonlinear optimization models while considering the characteristics of SDMWN and the capacity of wireless links to solve the CPP and select routes among network devices and controllers in the network. The results demonstrate the impact of different factors such as the number of controllers, the capacity of wireless links and the arrival rate of new flows in devices on control overhead, the average number of controller-device hops and the average number of inter-controller hops in SDMWN.

show abstract