الگوریتم salp swarm (ازدحام سالپ) بر پایه آشوب برای شبکه های چند کنترل کننده ای SDN

ABSTRACT Chaotic salp swarm algorithm for SDN multi-controller networks

Software Defined Networking (SDN) is a recent communication paradigm introduced for cost effective dynamic networks. The main idea behind SDN is the physical separation of the control plane and the forwarding plane through programmable software controllers, which enables dynamic configuration and control of the whole network [1]. This process differs completely from the traditional networks, in which the data plane is responsible for the whole process of data forwarding and thus both planes are completely integrated in the same device [2]. Data plane includes all forwarding devices that are responsible for traffic forwarding through the network, while the control tire contains all devices used for making traffic handling decisions. SDN approach covers the interaction and collaboration between the control and data planes, as the control plane controls and manages all forwarding devices in the data plane [3]. Control plane is responsible for setting configuration parameters and deciding the forwarding roles of all forwarding devices that perform traffic forwarding based on the received instructions. A set of intelligent controllers can be deployed in the control plane to act as the brain of the SDN network [4]. The command signals between the control and data planes are accomplished by means of an appropriate standard interface protocol (e.g. ForCES and OpenFlow) [5,6].

With the recent developments and advances in the industry of electronics and sensory manufacturing, the number of wireless devices is dramatically increasing. This puts high constraints on the design and development of the future systems (e.g. 5G, IoT and Tactile Internet) that cover the connectivity among these devices [7]. These design challenges include traffic volume, connectivity and reliability [8]. By 2020, it is expected that the data traffic will be 200 times higher than that in 2010 and by 2030 it will be 20,000 times higher than the traffic in 2010 [9]. Another important issue is the high requirements and demands required from the future systems [10]. The fifth generation of cellular system (5G) is expected to achieve an increased data rate and user demands up to 1000 fold (i.e. data rate of up to ten Gbps) with ultra low latency, high reliability, high mobility, high throughput and high connectivity [11,12].

This ultra high amount of traffic can’t be handled by current wireless solutions and thus, new technologies and paradigms should be involved. In order to achieve the requirements and user demands and handle the dramatic number of data traffic, 5G and next generation systems will deploy SDN and other technologies such as Mobile Edge Computing (MEC) and Network Function Virtualization (NFV) [13,14]. SDN network achieves various benefits through the decoupling of control plane and data plane. These benefits include [15,16]:

1.  Simplicity of network management,
2. Hardware simplicity,
3. Higher network flexibility,
4. Reduction of round trip latency,
5. High reliability,
6. The ability of network innovations, and
7.  Increasing overall system efficiency in terms of utilization [17].

Physical SDN controller can practically handle limited number of flow requests at a given time. Thus, the controller can only serve for limited number of switches and when the traffic increases above the upper limit that a controller can handle, there should be another controller to support. This indicates that the single centralized SDN controller may be prober, only for small scale networks [18]. The most common existing SDN interface is the OpenFlow, which is practically developed for single centralized controller [19]. OpenFlow protocol achieves poor efficiency, in terms of scalability, for large scale networks [20]. However, this problem can be improved by separating OpenFlow network into several fields, with one controller for each field. This separation and involvement of multiple controllers achieve the load balancing and scalability benefits to the SDN network [21]. Consequently, for large scale networks, employing single controller to manage all network switches isn’t efficient and thus, the efficient solution is to employ multiple controllers. Fig. 1 illustrates the basic architecture of the SDN multi-controller networks [22]. SDN networks, either single or multiple controllers, consist of three main layers; the data plane layer, the control plane layer and the application layer [1,23].