Multi-objective optimization controller placement problem in internet-oriented software defined network

doi:10.1016/j.comcom.2018.04.008

Computer Communications

Volume 123, June 2018, Pages 24-35

https://doi.org/10.1016/j.comcom.2018.04.008 Get rights and content

Abstract

In Internet-oriented Software Defined Network (SDN), the feasibility and scalability are guaranteed by the cooperation among multiple controllers, thus the effective placement of these controllers becomes the significant task. In this paper, we formulate a Multi-objective Optimization Controller Placement (MOCP) problem, which focuses on achieving high network reliability, load balance among controllers, and low latency among controllers and switches. The MOCP covers three specific questions: where the controllers should be placed, how to assign controllers to switches, and how many routing requests should be processed by each controller. We use Adaptive Bacterial Foraging Optimization (ABFO) algorithm and redefine its computation rules to solve the MOCP problem. Simulation results demonstrate that the proposed ABFO based MOCP scheme is effective and has better performance than the existing ones.

Introduction

Software Defined Network (SDN) [1] is a new networking paradigm which emphasizes the decoupling of control plane and data plane, realizes the user customized demands by programmability. SDN moves control logic off packet forwarding hardware, the complex functionalities in network devices are delivered to the specialized SDN controller. As a core component to realize SDN, SDN controller is responsible for maintaining the global view of network and providing the programmatic interfaces to the entire network [2], [3].

A single controller is often sufficient to meet existing response time requirements in many medium-size networks, but only one controller is not enough to manage the whole network with the expansion of network [4]. Furthermore, the physically centralized controller easily causes the single point failure over network, especially the large-scale or hyper-scale network, which significantly limits the reliability of network. If the controller breakdowns or becomes the performance bottleneck, the superiority of SDN will be lost [5]. Therefore, multiple controllers should be deployed and used cooperatively to improve network scalability, provide fault tolerance, and reduce latency of control plane. If one switch is connected to multiple controllers, these controllers can work together to handle routing requests from the switch. Although there are many studies have designed architecture of distributed SDN controller to enhance network scalability [6], [7], [8], [9], it is not efficient to increase the number of controllers merely. Both the number of controllers and their placement locations can affect the network performance significantly, thus it is necessary to investigate the controller placement problem [10]. That is to determine the layout and architecture of control plane, the optimal number of controllers, their locations in network topology, the optimal assignment of controllers to switches, and the optimal workload division among controllers. Heller et al. [10] formally presented the placement problem of controllers in the wide-area networks where the long latency between controllers or controller-switch leads to long response time, and then influences their ability to respond to network events.

In fact, the placement strategy of controllers has many influences on SDN performance [11], such as network reliability, operation cost and response time of events. A controller’s failure affects connection between control plane and forwarding plane, breaks down its controlled switches, even disables some functions of control plane. Hence, the controller placement strategy must take reliability into account to improve network availability. When the load of controller exceeds its processing capability, it will suffer more failures because it has no enough resources to deal with all arrived requests from switches [12]. Sometimes this may even lead to the cascading controller failure and increase the overhead of communicating with switches or other controllers, thus how to balance load among multiple controllers is vital. The controller processing latency will substantially increase with the increasing of flows from switches. There are many limitations on network availability, for example, whether controllers can response to network events or push forwarding commands to switches in time [13]. Therefore the latency constraints should be considered in Internet-oriented SDN.

A Multi-objective Optimization Controller Placement (MOCP) problem considering the above three optimization objectives is formulated in this paper. We use the Adaptive Bacterial Foraging Optimization (ABFO) algorithm to solve it and propose an ABFO based MOCP scheme. The major contributions of this paper are summarized as follows.

•
We formulate the MOCP problem, and the proposed solution can improve network reliability, balance load among controllers and reduce latency between controller and switch comprehensively.
•
The proposed scheme can select controller locations, determine mapping relationship among controllers and switches, and decide the proportion of routing requests from switches processed by each controller. In particular, there is many-to-many mapping relationship of controllers and switches, and the controllers cooperate on processing routing requests from the assigned switches.
•
We redefine the computation rules of ABFO, especially we customize its chemotaxis step to address MOCP. A 3-dimensional chemotaxis operation is devised to carry out 3 kinds of changes during each iteration time, i.e., change of the controller locations, change of the mapping relationship among controllers and switches, and change of the proportion of routing requests from switches processed by the controllers.

The rest of this paper is organized as follows. Section 2 reviews the related work. Section 3 describes the control network and then formulates the MOCP problem to be solved. Section 4 elaborates the controller placement scheme based on ABFO. Section 5 simulates the proposed scheme and evaluates its performance. Finally, this paper is concluded in Section 6.

Section snippets

Related work

There are many researches on SDN controller placement problem. Some of them are about control plane architecture design. In [6] and [14], a logically centralized but physically distributed control plane was proposed, switches reduced the cost of looking up flow by communicating with local controllers. In [7], a kind of hybrid control mode was proposed, the local applications employed local controllers while some policies needed centralized network status and would be redirected to global

Control network

The SDN control network is a logical network which overlays the physical network, and it consists of controllers and switches which are placed at the same location. In this paper, the SDN control network can be divided into multiple domains, and each one is controlled by some controllers. As depicted in Fig. 1, the SDN control network has two kinds of paths, one is general switch-switch connection, and the other one is control path, i.e., switch-controller connection and inter-controller

Expression of solution

In this paper, a solution of controller placement problem is denoted by a two-dimensional matrix $P = {[p_{i j}]}_{n \times (n + 1)}$ .

$P = [\begin{matrix} p_{10} & p_{11} & p_{12} & \dots & \dots & p_{1, n - 1} & p_{1 n} \\ p_{20} & p_{21} & p_{22} & \dots & \dots & p_{2, n - 1} & p_{2 n} \\ ⋮ & ⋮ & ⋮ & ⋮ & ⋮ \\ p_{i 0} & p_{i 1} & \dots & p_{i j} & \dots & p_{i, n - 1} & p_{i, n} \\ ⋮ & ⋮ & ⋮ & ⋮ & ⋮ \\ p_{n - 1, 0} & p_{n - 1, 1} & p_{n - 1, 2} & \dots & \dots & p_{n - 1, n - 1} & p_{n - 1, n} \\ p_{n 0} & p_{n 1} & p_{n 2} & \dots & \dots & p_{n, n - 1} & p_{n n} \end{matrix}]$

From this matrix, we can see where controllers are placed, which switches are controlled by which controllers and how many routing requests of switches are dealt with by each controller. There are n nodes in the network, $i = 1, 2, \dots, n$ means the node locations,

Evaluation and analysis

The proposed scheme is simulated by MATLAB and its performances are evaluated over three topologies which are Internet2 OS3E, Tinet from Internet Topology Zoo [38] and RF-II which is an ISP topology from the RocketFuel repository [39]. Internet2 OS3E has 34 nodes and 42 edges, Tinet has 53 nodes and 52 edges, and RF-II has 108 nodes and 306 edges. In this section, we analyze and evaluate our proposed ABFO based MOCP scheme with Heuristic algorithm in [27] and Pareto Simulated Annealing (PSA) in

Conclusion

The controller placement problem is to determine the layout of control plane, the optimal number of controllers to be placed and their locations in the network, etc. We define the MOCP problem and establish the mathematical model with the objective function aiming at improving control network reliability, balancing controller load and reducing control path latency to solve the formulated multiple controllers placement problem. We propose the method that combines heuristic rules and bio-inspired

Acknowledgements

This work is supported by the National Natural Science Foundation of China under Grant No. 61572123; the Major International (Regional) Joint Research Project of NSFC under Grant No. 71620107003; the National Science Foundation for Distinguished Young Scholars of China under Grant No. 71325002; the Program for Liaoning Innovative Research Term in University under Grant No. LT2016007.

References (40)

B. Yi et al.
A comprehensive survey of network function virtualization
Comput. Netw.
(2018)
J. Xie et al.
Control plane of software defined networks: a survey
Comput. Commun.
(2015)
B. Yi et al.
Design and evaluation of schemes for provisioning service function chain with function scalability
J. Netw. Comput. Appl.
(2017)
Y. Zhang et al.
On resilience of split-architecture networks
Global Communications Conference(GLOBECOM)
(2011)
T. Drezner et al.
Solving the multiple competitive facilities location problem
Eur. J. Oper. Res.
(2002)
J. Derrac et al.
A practical tutorial on the use ofnonparametric statistical tests as a methodology for comparing evolutionaryand swarm intelligence algorithms
Swarm Evol. Comput.
(2011)
D. Meyer
The software defined networking research group
Internet Comput.
(2013)
Software-defined networking: the new norm for networks, 2012, 4....
Y. Jarraya et al.
A survey and a layered taxonomy of software-defined networking
IEEE Commun. Surv. Tut.
(2014)
T. Koponen et al.
Onix: a distributed control platform for large-scale production networks
Proceedings of the 9th USENIX Conference on Operating Systems Design and Implementation
(2011)

S.H. Yeganeh et al.

A framework for efficient and scalable offloading of control applications

Proceedings of the First Workshop on Hot Topics in Software Defined Networks

(2012)

F. Yonghong et al.

A dormant multi-controller model for software defined networking

Communications

(2014)

A. Koshibe et al.

Towards distributed hierarchical SDN control plane

Science and Technology Conference(Modern Networking Technologies, MoNeTeC)

(2014)

B. Heller et al.

The controller placement problem

ACM SIGCOMM Comput. Commun. Rev.

(2012)

Open networking foundation, openflow switch specification (version 1.3.0),...

B. Yi et al.

A Multi-stage Solution for NFV-enabled Multicast over the Hybrid Infrastructure

IEEE Commun. Lett.

(2017)

Node.js,...

Devolved control of ATM networks,...

H. Yannan et al.

On reliability-optimized controller placement for SDN

China Commun.

(2014)

Y. Zhang et al.

Fast failover for control traffic in software-defined networks

IEEE Global Communications Conference(GLOBECOM)

(2012)

Cited by (65)

A novel framework for capacitated SDN controller placement: Balancing latency and reliability with PSO algorithm
2024, Alexandria Engineering Journal
Software Defined Networking (SDN) is a new network architecture theoretical framework, which based on the partition of the control plane and the data plane. Some of the important features of SDN are centralized network management, open code, scalability, effective utilization of network resources, effective traffic management, security etc. All these administrative decisions taken by the controller which lies at the top of the data plane. Efficient placing of controllers is essential to enhance the network’s reliability, which ultimately improves the performance of the network. Much of the work done in the SDN has been aimed on controller placement problems (CPP) by considering latency and reliability of the network separately whereas few of them acknowledged both latency and reliability simultaneously. We considered both communication latency and reliable communication to solve CPP. To achieve this, we aim to minimize the total average latency of the network with a reliable network model that can work perfectly up to the failure of n-1 controllers out of n deployed. We first formulate an efficient algorithm “Capacitated Controllers Arrangement (CCA)” to define the problem and used nature-based metaheuristic technique, Particle Swarm optimization (PSO) to achieve the solution. By intelligently allocating controllers, this dynamic system minimizes network latency and optimizes switch assignments. Three controllers deliver excellent reliability and latency performance even despite controller breakdowns, according to experiments conducted on the Internet2 OS3E architecture. This novel method not only increases SDN efficiency but also establishes a standard for further improvements in controller placement tactics. The outcome shows that the proposed approach achieves its goal efficiently and in future other multi-objective optimization techniques will be explored to solve reliability aware CPP.
Autonomous solution for Controller Placement Problem of Software-Defined Networking using MuZero based intelligent agents
2023, Journal of King Saud University - Computer and Information Sciences
Controller placement is a critical issue in Software-Defined Networking, which has been identified as a potential approach to achieve more flexible network control and management. Deep Reinforcement Learning holds immense promise in achieving favorable outcomes through its ability to explore solution spaces and adapt to swiftly changing data streams. Furthermore, the advancement of reinforcement learning algorithms, like chess or Go, has inspired the concept of treating controller placement problem as a Markov game. These algorithms are employed to train intelligent agents capable of autonomously resolving this problem. This paper presents an intelligent system capable of optimizing the placement of controllers in Software-Defined Networking. The Muzero Reinforcement learning algorithm is used to train a model via self-competition which, by combining a tree search with a learned model, achieves superhuman performance in complex domains. Once trained, the model is integrated with an Software-Defined Networking controller, so that it is able to find the optimal placement of the game in a real network by incorporating performance metrics including latency, load, and overhead communication into the training process of intelligent agents. In order to show that our approach is feasible and efficient in practice, experiment results are provided as a benchmark.
Placement of SDN controllers based on network setup cost and latency of control packets
2023, Computer Communications
Software-defined networking (SDN) was developed to simplify the management of computer networks. In the SDN architecture, the control layer is completely decoupled from the data layer. Moreover, network software configuration is also possible. To setup an SDN, you need a device called controller, which is not available in traditional networks. In the control layer, there are several controllers that control the data layer switches. The type, number and place of these controllers directly affect the cost and average latency of control packets. Since achieving the lowest setup cost and the minimum latency of the control packets (between switches and controllers, and among controllers) is of great importance at the same time, in this paper, a binary linear programming model is proposed to achieve the trade-off between them in wide area networks(WAN). The model determines the type, place and minimum number of required controllers by receiving the place of the switches, the amount of processing capacity of the controllers and the setup cost of each controller. Accordingly, it reduces the setup cost, average latency between switches and controllers as well as average latency among controllers in comparison with previous methods. The results of the study show that the model has the lowest setup cost and the minimum average latency of control packets (switch-to-controller and controller-to-controller) in controllers placement compared to the previous methods.
Effective multi-controller management and adaptive service deployment strategy in multi-access edge computing environment
2023, Ad Hoc Networks
Citation Excerpt :
The relevant services are migrated to the backup server when a server goes down. Zhang et al. [17] propose a combinatorial optimization approach that guarantees that services can be unaffected when a certain number of hosts fail. However, the number of host failures is certain.
With the popularity of emerging mobile services such as virtual reality and online gaming, users expect higher quality immersive experiences. In response to the communication reliability and latency performance problems between controllers and switches due to single link failure in the network, a framework for multi-access edge computing systems based on software-defined networking (SDN) has emerged, and this paper proposes a reliable placement method for controllers based on deployment cost and link failure probability. Reliability and worst-case propagation delay model based on link failure probability is constructed. A controller placement strategy with minimum network cost is solved by a controller layout algorithm based on density-peak clustering. A mobile-aware service adaptive deployment method is also proposed in this paper to address the problems of long user-perceived delays and high migration costs due to the limitations of edge server resources and coverage areas and the dynamic nature of user movement. A mobility model based on the probability density function of sojourn time and a service deployment overhead model based on user mobility are constructed. A quantum ant colony algorithm is used to solve the service adaptive deployment scheme. The proposed algorithm is compared with the benchmark algorithm using the vehicle detection benchmark application service. Experimental results show that the proposed controller placement algorithm weighs the deployment cost, worst-case propagation delay and reliability of controllers and gives a reasonable number of controllers and their locations. The proposed deployment algorithm can optimise the quality of the user experience.
Qos-aware mobile service optimization in multi-access mobile edge computing environments
2022, Pervasive and Mobile Computing
With the rapid development of mobile Internet technologies and various new service services such as virtual reality (VR) and augmented reality (AR), users’ demand for network quality of service (QoS) is getting higher and higher. To solve the problems of high load and low latency in-network services, this paper proposes a data caching strategy based on a multi-access mobile edge computing environment. Based on the MEC collaborative caching framework, an SDN controller is introduced into the MEC collaborative caching framework, a joint cache optimization mechanism based on data caching and computational migration is constructed, and the user-perceived time-lengthening problem in the data caching strategy is solved by a joint optimization algorithm based on an improved heuristic genetic algorithm and simulated annealing. Meanwhile, this paper proposes a multi-base station collaboration-based service optimization strategy to solve the problem of collaboration of computation and storage resources due to multiple mobile terminals and multiple smart base stations. For the problem that the application service demand in MEC server changes due to time, space, requests and other privacy, an application service optimization algorithm based on the Markov chain of service popularity is constructed, and a deep deterministic strategy (DDP) based on deep reinforcement learning is also used to minimize the average delay of computation tasks in the cluster while ensuring the energy consumption of MEC server, which improves the accuracy of application service cache updates in the system as well as reducing the complexity of service updates. The experimental results show that the proposed data caching algorithm weighs the cache space of user devices, the average transfer latency of acquiring data resources is effectively reduced, and the proposed service optimization algorithm can improve the quality of user experience.
A cross entropy based approach to minimum propagation latency for controller placement in Software Defined Network
2022, Computer Communications
Citation Excerpt :
Fourth, the controller placement demonstrations are plotted in MATLAB. In order to evaluate the performance of the proposed approaches, we compare the cross entropy with four representative solutions in the literature: K-means, K-means++, Adaptive Bacterial Foraging Optimization (ABFO) [24] and Pareto Simulated Annealing (PSA) [25]. Moreover, the experimental results of the cross entropy are further compared with the optimal solution.
In recent years, the Software Defined Network (SDN) has emerged as a pivotal element not only in data-centers and wide-area networks, but also in next generation networking architectures. SDN is characterized by decoupled data and control planes with logically centralized architecture. In order to span across the networks and avoid single point of failure, one major challenge in the SDN is to select appropriate locations for controllers to shorten the latency between controllers and switches, especially in wide area networks. For this purpose, we formulate the Controller Placement Problem (CPP) as an integer programming problem, which takes both the communication cost and synchronization cost into account. Due to its high computational complexity, the cross entropy belonging to the field of Stochastic Optimization is proposed and can sample the problem space and approximate the distribution of good solutions. As a result, we propose a cross entropy based approach to solve CPP, and we conduct experiments on 6 real topologies from the Internet Topology Zoo and Internet2 OS3E. The results verify that the proposed approach can realize the minimum propagation latency for different network scales with different number of controllers, with a less than 5.30% margin from the optimal solution. Moreover, the cross entropy can promise the calculation result be stable with a less than 2% margin, and can apply to all the network scales including large network topologies.

View all citing articles on Scopus

View full text

Multi-objective optimization controller placement problem in internet-oriented software defined network

Abstract

Introduction

Section snippets

Related work

Control network

Expression of solution

Evaluation and analysis

Conclusion

Acknowledgements

Comput. Netw.

Comput. Commun.

J. Netw. Comput. Appl.

Eur. J. Oper. Res.

Swarm Evol. Comput.

The software defined networking research group

Internet Comput.

A survey and a layered taxonomy of software-defined networking

IEEE Commun. Surv. Tut.

Onix: a distributed control platform for large-scale production networks

Proceedings of the 9th USENIX Conference on Operating Systems Design and Implementation

A framework for efficient and scalable offloading of control applications

Proceedings of the First Workshop on Hot Topics in Software Defined Networks

A dormant multi-controller model for software defined networking

Communications

Towards distributed hierarchical SDN control plane

Science and Technology Conference(Modern Networking Technologies, MoNeTeC)

The controller placement problem

ACM SIGCOMM Comput. Commun. Rev.

A Multi-stage Solution for NFV-enabled Multicast over the Hybrid Infrastructure

IEEE Commun. Lett.

On reliability-optimized controller placement for SDN

China Commun.

Fast failover for control traffic in software-defined networks

IEEE Global Communications Conference(GLOBECOM)