12.1 Introduction

Software Defined Network (SDN) [1] is turning into an inventive worldview for cutting edge systems. The characterizing highlight of SDNs contrasted with customary systems differentiates the system control plane from the information plane, improves arrangement of the board, and enables it as increasingly adaptable. It makes the system to shield the basic unpredictability and controller gives a basic with the effective setup of the application layer at the upper level [2]. Controller decides the sending way for each stream in the system. Open Flow [3] is an institutionalized medium which can speak with information plane. Open Flow isolates control capacities from the system gadgets and keeps up the data stream monitoring table on the system gadgets. Segment Routing or SR can streamline the system abilities of IP, MPLS as well and empower the system to accomplish better adaptability. The key thought of SR is to perform directing by a couple of centres focuses dependent on a succession of consistent sections shaped between the entrance and departure hubs. What’s more, the moderate change just has to realize how to arrive at the center focuses and to advance bundles.

SR [5] maintains a strategic distance from the prerequisite for many label encodings to be put away along every way in each system gadget, so the SR diminishes the quantity of sending rules in TCAM [6]. Furthermore, SR wipes out the multifaceted nature of keeping up a mass of sending rules, it can result the low computation overhead in case of traffic engineering [7, 8]. In the administrations, clients or systems have various prerequisites. For instance, a few clients have prerequisites of giving free ways to accomplish steering division in their systems. A few clients need to keep away from exceptional ways in directing, and some cost heartless clients require the base deferral or higher transfer speed. The blend of SR with SDN can perform well with the separated system administrations and immediately set up and actualize organization. Traffic handling is important to improve the productivity of the system execution for administrators. The controller doesn’t have to add additional sending rules to every individual switch along with the path information [9]. It just needs to monitor the data from the start to finish of the data transmission. In this way, SR enormously improves the intricacy of system hardware. The middle of the road hubs just need to advance the IP bundles as per the arranged records in the section header [4]. All the state data of steering is kept up by the entrance hub.

Several network protocols have included support for some forms of source routing. IPv4 supports a source routing option that allows a host to indicate intermediate hops on the path towards a given destination. Most IPv4 implementations support this form of source routing. However, it is rarely enabled because of several security problems [2], and because end hosts have difficulties in obtaining accurate information about the path towards a given destination. Source routing was also a feature of the original IPv6 specification. IPv6 defines the Routing Header type 0 to encode a source routed path for a packet. IPv6 faced the same problems as IPv4 for the deployment and utilization of source routing beyond labs and simple networks. From a security viewpoint, IPv6 did not solve the security problems that affected IPv4 source routing and serious attacks against IPv6 source routing were published [1]. Given the potential impact of these attacks, the Internet Engineering Task Force (IETF) eventually decided to completely deprecate the IPv6 type 0 Routing Header [1]. Despite this, the ability to indicate inside each packet a set of intermediate hops remains a very useful feature to solve a wide range of networking problems [3, 4]. Given the importance of these problems, network vendors and network operators convinced the IETF to create the Source Packet Routing in Networking (SPRING) working group to standardize a modern source routing solution. The first approach discussed within the SPRING working group leverages the Multiprotocol Label Switching (MPLS) data plane. In a network using MPLS, routers receive information about the topology through the intra domain routing protocol (OSPF or IS–IS). Extensions to these protocols have been proposed to use them to distribute MPLS labels associated with routers or specific links. With this 1Given the restriction on the number of citations, we do not reference all IETF documents.

A bibliography on Segment Routing is maintained at http://www.segment-routing.net information, where any router in the MPLS network can select a source routed path towards any other router inside the network and encode it as a label stack which is attached to each packet. Various use cases are being defined and implemented with this MPLS data plane in service providers networks. The second approach pursued within the SPRING working group is to design a source routing extension for IPv6 [8]. IPv6 has a much broader applicability in the long term than MPLS which is restricted to service providers and some enterprise or data center networks. However, with IPv6 we cannot simply assume as in MPLS that only trusted routers will generate source routed packets. Any IPv6 node can generate source routed packets and a modern source routing extension for IPv6 must take security into account. In this chapter, we explain the main design principles that underlie the IPv6 Segment Routing extension header and report our experience with the first complete open-source implementation of this technology [9].

12.2 Software-Defined Network

Software Defined Network or SDN is a kind of network that enables the network to become intelligent and automated with a central authority which can be either of applications or a kind of software. The SDN makes the network administrator capable of handling the traffic and the whole network consistently without bothering the adopted network topology. The control area of the SDN is basically handled by some dedicated API. The SDN can be further defined with the four columns such as:

1. 1) Both of the control and information planes are distinguished [14].
2. 2) Sending packets is stream-based, rather than goal-based. A data packet is defined by the number of datums that are intended for some of the devices within the network. With the data packets the path will be ensembled in order to find a justified destination. Upon arriving at the destination the path history is being checked and verified that the packet has arrived at proper place. A single control plane handles all of the input traffic [25, 26]. The stream deliberation permits bringing together the conduct of various sorts of system gadgets, including switches, switches, firewalls, and middle boxes [27].
3. 3) Control rationale is moved to an outside element, the so-called SDN controller or System Working Framework which is a product stage that suddenly increases the demand of the data from the devices present in the network. The SDN provides all basic assets to handle the heavy amount of traffic with in the network. The control plane is being handled by some dedicated API which is user friendly in nature [10, 11].
4. 4) SDN has been designed in such way that working framework can elaborate some basic information plane assets. The working framework has been capable for handling and monitoring the traffics coming from various nodes present in the network which becomes the motivation to adopt the SDN [13].

To start with, it is less complex and what’s more, less mistake inclined to adjust arrange strategies through high-level dialects and programming segments, contrasted and low-level gadget explicit designs. Secondly, it can consequently respond to deceptive modifications of the system state and in this way keep up the elevated level strategies flawless [20]. Last but not the least, centralizing the controlling rationale in a controller with worldwide information on the system state improves the advancement of progressively refined systems administration capacities, administrations, and applications Figure 12.1. Following the SDN idea presented in [21], a SDN is further characterized by three key reflections:

1. (i) Transferring the data in the network
2. (ii) Distributing the appropriate data
3. (iii) Specifying the accurate source and destination.

A) Elements of SDN: There are various elements present in the SDN that can be represented as follows [28]:

• Forwarding Devices (FD)
• Data Plane
• Southbound and Northbound Interface
• Control and Management Plane.

Forwarding Devices: The forwarding devices such as hardware devices and software APIs execute the computation process. The interfaces such as Openflow, ForCES, and POF are being installed in the forwarding devices by the controller [29].

Data Plane: The FDs are connected with each other through the wireless or wired connection. The data plane is nothing but the network of interconnected devices or the FDs [30].

Southbound and Northbound Interface: The set of instructions provided by the FDs is being defined by the southbound interface. The NOS provides the northbound interface as API to the application developer. Basically, the northbound interface abstracts the low-level instructions sent by the southbound interface [30].

Control and Management Plane: The control plane acts as a core part of the network which controls all types of instructions. Whereas the management plane is responsible for the activity such as the routing, firewalls, and the load balancer etc. Basically, an administration application characterizes the arrangements, which are at last meant southbound-explicit guidelines that program the conduct of the sending gadgets [31].

12.3 SDN Architecture

A SDN can be delineated as an arrangement of various layers (Figure 12.2). Each layer possesses its explicit capacities. While some of them are constantly present in a SDN arrangement, for example, the southbound Programming interface, organize working frameworks, northbound Programming interface, and system applications, others might be available just specifically arrangements, for example, hypervisor-or language-based virtualization [32].

An SDN foundation, comparatively to a customary system, is made out of a lot of systems administration hardware (switches, middleware machines) [33]. The fundamental contrast dwells in the actuality that those conventional physical gadgets are presently straightforward sending components without installed control or programming to take independent choices [34]. The system insight is expelled from the information plane gadgets to a legitimately brought together control framework, i.e., the system working framework and applications. All the more critically, these new systems are assembled (thoughtfully) on open furthermore, standard interfaces (e.g., OpenFlow), a critical methodology for guaranteeing setup and correspondence similarity furthermore, interoperability among various information and control plane gadgets [35]. At the end of the day, these open interfaces empower controller substances to progressively program heterogeneous sending gadgets, something troublesome in customary systems, due to the enormous assortment of exclusive and shut interfaces and the dispersed nature of the control plane.

In the SDN there are two basic elements present known as the controller and the sending elements. The information plane device is kind of device which is being programmed for handling the network traffic [36] whereas the controller is responsible for controlling everything the network including the monitoring the data packet that has been transmitted with the network. The SR with the SDN empowers the sending and receiving devices to be remain in active condition with the presence of huge amount of traffic. Each message in SR-SDN will have three different sections such as the path information, activity information and the counters to reduce the overhead of the node [37].

Southbound interfaces (or southbound APIs) are the associating spans among control and sending components, in this way being the pivotal instrument for obviously isolating control and information plane usefulness [40, 41]. Notwithstanding, these APIs are still firmly attached to the sending components of the hidden physical or virtual foundation. Regularly, another switch can take two years to be prepared for commercialization whenever worked without any preparation, with overhaul cycles that can take as long as nine months. The product improvement for another item can take from a half year to one year. The beginning speculation is high and dangerous. As a focal segment of its structure, the southbound APIs speak to one of the significant hindrances for the presentation and acknowledgment of any new organizing innovation. Right now, the development of SDN southbound Programming interface proposition, for example, OpenFlow [38] is viewed as welcome by numerous individuals in the business. These measures advance interoperability, permitting the sending of merchant rationalist arrange gadgets. This has just been exhibited by the interoperability between OpenFlow-empowered hardware from various merchants [39].

Virtualization is as of now a combined innovation in present-day PCs. The quick advancements of the previous decade have made virtualization of registering stages standard. Based on late reports, the quantity of virtual servers has as of now surpassed the number of physical servers. Hypervisors empower unmistakable virtual machines to share the same equipment assets. In a cloud IaaS, every client can have its own virtual assets, from figuring to capacity. This empowered new income and business models where clients dispense assets on request, from a mutual physical foundation, at a generally minimal effort. At the same time, suppliers utilize the limit of their introduced physical foundations, making new income streams without altogether expanding their CAPEX and OPEX costs [42– 45]. One of the fascinating highlights of virtualization advances today is the way that virtual machines can be handily relocated starting with one physical server then onto the next and can be made and additionally pulverized on-request, empowering the provisioning of flexible administrations with adaptable and simple administration [46]. Lamentably, virtualization has been just mostly acknowledged practically speaking. In spite of the incredible advances in virtualizing figuring and capacity components, the system is still for the most part statically arranged in a box-by-box way. The general OS gives the high level APIs to manage the simultaneous access to the available resources such as CPU, memory etc along with the security mechanism to protect those from integrity loss [47]. These characteristics are playing vital role in productivity increment making the process more reliable in developing and managing the applications. The wide use of programming language helps in evolution of the various application systems.

12.4 Segment Routing

Dissimilar to RSVP and Label Distribution Protocol (LDP), SR requires no MPLS control plane flagging and forces no progressions to the MPLS information plane. SR requires just entrance name edge switches to keep the per-administration state. State the executive’s prerequisites from the midpoint (name switch switches) and the last part (departure mark edge switches) are evacuated. This permits SR to scale essentially superior to RSVP-TE while giving the vast majority of similar capacities: Interior Door Convention (IGP)-based MPLS burrows for administrations [48], for example, VPRN or VPLS with no other vehicle flagging convention. Fast-reroute capacity utilizing a pre-processed reinforcement way that can give full inclusion and doesn’t have any topology conditions. Ability to source course utilizing a mix of free as well as exacting bounces. Takes into consideration brought together or disseminated traffic building models with the greater part of the capacities of RSVP-TE (counting administrator gatherings and shared hazard interface gatherings), without the related midpoint state the executive’s necessity that causes scaling issues in bigger systems [49]. While SR bolsters appropriated traffic designing, choices to improve and guarantee set up mark exchanged ways (LSPs) are constrained because of the absence of state data on executed courses. A PCE tends to this issue by keeping up arrange wide SR topology and per-LSP state in a brought together rush hour gridlock building database (TED) [40].

A PCE server can either be stateful or stateless. A stateful PCE server has extreme synchronization between the PCE and the framework state using the TED and keeps up state on the course of action of dynamic ways and their spared resources in the framework. A stateless PCE server registers ways reliant on the TED yet frames each way self-sufficiently and doesn’t need to review any as of late figured ways. It isn’t plausible for a stateless PCE server to process improved unique ways due to the nonappearance of LSP states [40]. The use of a stateful PCE server improves the believability of a perfect way count yet requires a strong synchronization part among itself and the framework. A PCE server can either be stateful or stateless. A stateful PCE server has severe synchronization between the PCE and the system state utilizing the TED and keeps up state on the arrangement of dynamic ways and their saved assets in the system. A stateless PCE server figures ways dependent on the TED yet forms every way freely and doesn’t have to recall any recently processed ways. It isn’t workable for a stateless PCE server to figure advanced dynamic ways because of the nonattendance of LSP states [22]. Utilization of a stateful PCE server improves the plausibility of an ideal way calculation however requires a solid synchronization component among itself and the system. Discovery of system topology by tuning in to IGP/TE-asset refreshes through a course audience or through BGP interface, collects interface insights and acknowledges demands from arrange components for way calculation utilizing the Path Calculation Component Convention (PCCC), monitors arrange state, estimating such things as continuous (or close to ongoing) traffic request and LSP insights, and re-enhances way position dependent on that information, supports both RSVP and SR LSP types [23]. There are numerous favorable circumstances to applying portion directing innovation into transporter Ethernet systems. Here are two significant points of interest: Simplicity: As it appears in the figure, fragment steering can expel the requirement for MPLS LDP, LDP FRR, RSVP-TE, and TE-FRR, RFC3107 BGP transport [50].

12.5 Segment Routing in SDN

SR [4] is an as of late proposed steering engineering to take care of the perplexing issue of an enormous number of sending rules in the data sending process. A portion of [51] speaks of guidance which is utilized to characterize the way in a weighted diagram. A hub fragment contains the novel name of the following change to reach. The halfway change just has to realize how to advance to the following moderate hub in the briefest way. In the present draft characterized by the IETF, two portions are characterized: the Nodal Fragment and Nearness Section. The Nodal Portion is a worldwide name [52]. Every hub is allowed by an all-around interesting Nodal Fragment. The Nearness Fragment is a nearby mark that is locally legitimate and used to speak to a specific section steering hub [45]. The SDN controller computes the unequivocal course for the steering module and arranges the sending table of the entrance switch with a fragment list [50].

Shown in Figure 12.3, is a simple network which describes the segment routing based on SDN network. The SDN controller first calculates the path from source to destination and then forwards the packets to the corresponding router. Initially, the SDN controller sends the packet to hub A. Each node uses an upper-level stack with the intermediate nodes. A forwards the packet while mentioning the adjacent intermediate node C with a packet of 105 in our experiment. The hub C pops up the top level of the stack and decides the next switch D to which the packet has to be forwarded and forwards it to the next node which can transfer the packet to the corresponding destination F. Similarly the switch E pops up the top level of the stack and finds that it can transfer the packet to the intendant recipient F and immediately transfers the data *packet to F. The segment routing maintains the route details which is being used for the communication from the source to destination F. The level stack is being initially created by the SDN controller. In this way, the SDN control plane acts to forward the packet from the sender to the receiver [53].

The authors in [4] represent the concepts of SR and its usage. SR utilizes the source directing technique. The possibility of SR is to partition a start to finish way into various portions, every one of which is distinguished by a fragment identifier. The sending steps are assigned in the fragment, so the switches don’t need to keep up steering table data. The header of the bundles conveys the section. The hub advances the parcel to the goal by means of the briefest way. The creators note that SR has an exceptional bit of leeway in actualizing system load parity and traffic designing. As we probably are aware, SDN has a worldwide perspective on the system and can control the portions. The authors in [18] target taking care of the issue of reclamation streamlining in SR. The fundamental issue is to bring together associated with the essential way to accomplish the ideal sharing of recuperation data transfer capacity in non-synchronous system disappointments. The creators propose a straight programming issue and an irregular adjusting plan right now. The base double calculation takes care of the direct programming issue, and it has a decent running time even on huge systems. The straightforward irregular adjusting plan focuses on different limitations on section steering. In any case, our exploration focuses on the assessment and determination of ways for SR in SDN. The authors in [19] focus on two related circumstances: dynamic traffic recuperation and multi-area traffic designing. This chapter characterizes the fitting strategy to limit the profundity of the section list. As to traffic recuperation, a SRFAILOVER plot is proposed to progressively recoup traffic stream hindered by connection or hub disappointments to limit the profundity of the necessary portion list. What’s more, most reinforcement ways can be encoded with a fragment rundown of a couple of marks. The shortest path first (SPF) [12] calculation utilizes the briefest way calculation to compute the way.

The SPF calculation doesn’t consider if the transfer speed of the chose way fulfils the need of the real application. The possibility of the WSP [14] is to discover the way with the biggest outstanding transfer speed in all min-jump ways The possibility of MIRA [15] is to limit the effect on the foundation of connection demands for future hubs. The authors in [16] propose a steering calculation for SDN with SR. The calculation can meet the transmission capacity necessity of directing solicitations and decrease the additional expense of the bundle header size. In any case, the calculation doesn’t consider inclinations of clients to assess the connection execution of a system [17].

12.6 Traffic Engineering in SDN

Traffic building systems in SDN can be a lot all the more effectively and brilliantly executed as an incorporated TE framework contrasted with the traditional approaches, for example, ATM-, IP-, and MPLS-based TEs in light of the fact that of the significant preferences of the SDN engineering [57]. All the more explicitly, SDN gives (1) concentrated deceivability including worldwide system data (e.g., organize asset restrictions or progressively changing the system status) and worldwide application data (e.g., QoS necessities); (2) the programmability without taking care of person foundation components, i.e., OF switches at the information plane can be proactively customized and powerfully reconstructed by the brought together controller to ideally distribute organize assets for arranging blockage evasion, what’s more, upgraded QoS execution; (3) receptiveness, where information plane components (i.e., OF switches), paying little mind to the sellers, have a bound together interface open to the controller for information plane programming and system status assortment; also, (4) numerous stream table pipelines in OF switches can make stream the board progressively adaptable and proficient [54]. Since the development of SDN, it has been applied to an assortment of system conditions, (i.e., Undertaking systems, enormous scope server farm systems, WiFi/ cell systems, and so forth). TE innovation is of basic significance to the advancement and accomplishment of SDNs. To start with, as indicated by the essential activity of stream the board in SDNs, when a stream showing up at switch doesn’t coordinate any guidelines in the stream table, it will be handled as follows: (1) the primary bundle of the stream is sent by the entrance change to the controller, (2) the sending way for the stream is figured by the controller, (3) the controller sends the fitting sending sections to introduce in the stream tables at each switch along the arranged way, what’s more, (4) every consequent bundle in the stream or even unique streams with coordinating (or comparative) qualities are sent in the information plane along the way and needn’t bother with any control plane activity [56]. Right now, the collected traffic comprises of high number of new streams, a critical overhead can be yielded at both the control plane and information plane. Besides, the sending rule arrangement can likewise take time, with the goal that the inactivity can be expanded. Along these lines, to take care of these issues, traffic designing instruments for the stream the board ought to be intended to address the trade-offs between the dormancy and burden balance [55].

12.7 Segment Routing Protocol

A) Multi-Objective Particle Swarm Optimization: The traffic engineering in the network means handling the severe traffic during the communication. The main objective of traffic engineering is to analyze and optimize network performance. With the different kinds of network traffic engineering evolves different types of requirements. But the stability of the network and cost will be still a challenge in traffic engineering. The SDN provides a better exposure for the traffic engineering concept to deal with the stability and cost while handling the large amount of data. The PSO is an intelligent algorithm that optimizes through the collaboration of the birds. MOSPO focuses to solve the MOO problem by setting up various populations. The optimal solution is achieved by using various populations. The steps of MOSPO [1] is as follows

1. 1) The population size is being initialized with the customize.
2. 2) Define the corresponding objective function.
3. 3) Store the non-inferior solution by considering the Pareto dominance relationship.
4. 4) Choose the global optimal value.
5. 5) Recalculate the velocity and position of every particle with respect to objective function.
6. 6) Determine the fitness value.
7. 7) Reform optimum point for both local and global
8. 8) If the termination condition is not satisfied then iterate from step 3.

At the nth iteration Entropy E(n) is being calculated as the following:

(12.1)

C_k,m(n) = number of cells falling to the m column. The entropy is being used for performance calculation. For two continuous iteration, the ΔE(n) is calculated as

(12.2)

The global optimal solution is defined as

(12.3)

dbest(n) is the best global optimal solution if the entropy difference between two consecutive iterations is less than threshold entropy δ_c and cbest(n) is the best global solution if the entropy difference between two consecutive iterations is greater than threshold entropy δ_c.

B) Advance MOSPO

As SDN has a global perspective on the system asset, it can get the worldwide system traffic to plan a superior methodology for arrange improvement. With the presence of a given traffic request from the controller of the SDN, interface weight decides one of briefest way between source and destination hub present in the network. Also it impacts the way cost and burden dissemination in the system. Hence, the enhancement of connection weight has become a powerful method to acknowledge traffic building. Right now, utilize the way cost and burden parity of the system as the two streamlining destinations for the MOPSO calculation and improved MOPSO calculation. The advanced MOPSO [1] calculation for the advancement of connection weight is outlined in the following calculation.

C) Minimum Interference Routing Algorithm (MIRA): A system of n routers is considered. A subset of these routers is thought to be entrance egress routers between which LSPs can be set-up. Nonetheless, there doesn’t have to be a potential LSP between each entrance and each departure. Rather, from a specific entrance, LSPs might be passable as it were to specific departures [16, 24]. This might be a result of strategy or administration requirements, (for example, certain VPN traffic may just begin and exit at certain entrance departure sets). It is accepted that any such data is known, changes not every now and again, and is made accessible to the course server (we depict for straightforwardness just a concentrated course calculation in the chapter) by provisioning or on the other hand regulatory mechanism. Input to this MIRA algorithm is the graph G having n number of routers. A set of all residual link capacity of the graph is denoted by B. Any two ingressive nodes a and b in between which a data flow of D units has to be transferred.

1. 1. The maximum flow value for all possible ingressive pairs of nodes is being calculated.
2. 2. The critical link C_s,d is calculated.
3. 3. The weight for all critical link of s and d is calculated as follows:
   (12.4)
4. 4. The links having residual bandwidth less than D, then those links need to be eliminated.
5. 5. Use the Dijkstra algorithm shortest path in between the ingressive node is being calculated.
6. 6. Using the shortest path the packet of D units has to be routed from ingressive node a to b.

12.8 Simulation and Result

The authors have used the MATLAB and Wireshark for the simulation process. In the experiment, random nodes have been chosen as source and destination. Different paths and different links have different packet loss rates. The path loss rates for three different algorithms are mentioned in Table 12.1. Tables 12.2, 12.3 and 12.4 shows the average throughput, hop count and link utilization with different network size. The authors have taken six different paths during the experiment. A total of 50,100,150,200 number of nodes have been taken for the experiment.

12.9 Conclusion and Future Work

Traffic engineering is becoming the most challenging factor in a network where large number of communication present in the same network. But the stability and the cost of the network in terms of multi-hop communication. The SDN with the help of segment routing plays a vital role in handling the traffic while optimizing the link utilization and increasing the network performance. In this research work, the authors have tried to study the Software Defined Network and the impact of Segment Routing on it. Also, implemented three different segment routing MIRA, MOPSO, and AMOPSO over different network sizes. The authors have created Waxman Network Topology (WNT) with four different network sizes having the number of nodes 50, 100, 150 and 200 respectively. After implementing these above-said algorithms the authors have tried to calculate the network performance by considering some influencing parameters such as delay (Figure 12.4), path loss rate (Figure 12.5), energy consumption (Figure 12.6), load balancing (Figure 12.7), and maximum link utilization (Figure 12.8). By taking all of the parameters the advance MOPSO performs well while dealing with a large network. Figures 12.9, 12.10 and 12.11 shows the performance evaluation of MIRA, MOPSO and AMOPSO in contrast to average throughput, hop count and link utilization respectively. AMOPSO completely satisfies the main objective of the research i.e., optimal link utilization while communicating in between different nodes having various communication paths. Segment routing helps the SDN in determining the appropriate path among all of the available possible paths present in the WNT. In the future, the authors will try to implement the proposed algorithm by taking a larger network size. Moreover, the authors will consider some unique parameters and compare the performance of AMOPSO with the existing one, and further research will be done on actual networks by enchanting the rejection rate.