CU-MCA-SEM-I-Advanced Computer Networks-Second Draft

It aids in the identification of similar land areas for use in an earth observation database, as well as the classification of home groups in a city based on house type, value, and geographic position. Fig 14.2 Clustering 14.5 PURPOSE OF CLUSTERING Due to the wide range of applications, clustering analysis has become a developing challenge. The popularity of these methods is likely due to the introduction of new data clustering techniques in recent years and their widespread use in a variety of applications, such image processing, computational biology, mobile communication, medicine, and economics. The fundamental problem with data clustering methods is that they are non-standard. With one sort of data set, the advanced algorithm may produce the best results, but with another type of data set, this could fail or perform poorly. Despite several efforts to standardise algorithms which can perform in all contexts, no major progress has been accomplished thus far. So far, many clustering tools were proposed. However, every algorithm has its own set of benefits and drawbacks, and none of them can be used in all real-world scenarios. 1. Scalability is important. When it comes to clustering, scalability means that as the number of data items grows, the time it takes to cluster them should roughly scale with the algorithm's complexity order. For instance, we know that K-means clustering is O (n), where n represents the total of objects inside the data. If the amount of data objects is multiplied by ten, the time it takes the cluster 251 CU IDOL SELF LEARNING MATERIAL (SLM)

them should be multiplied by ten. It implies that the relationship should be linear. If this isn't the case, something went wrong during the implementation process. 2. Interpretability: Clustering results should be understandable, clear, and useful. 3. Identifying clusters based on attribute shape: Clustering algorithms should be able to detect clusters of any shape. They shouldn't be restricted to distance measures that reveal a spherical cluster of modest sizes. 4. The ability to deal with a variety of characteristics: Algorithms should be able to work with any type of data, including numeric data, binary data, or categorical data. 5. The ability to deal with a large amount of noisy data: Data which is noisy, missing, or erroneous might be found in databases. Few algorithms were sensitive to this type of data, which can lead to low-quality clusters. 6. A high degree of dimensionality: Clustering tools should be able to handle both high and low-dimensional data sets. 14.6 SUMMARY • Routing protocols make it easier for routers to learn routes by requiring each router to announce the routes it is aware of. • Each router starts with only the routes that are immediately connected to it. Then, each router transmits messages that list the routes, as described by the routing protocol. • When a router receives a routing help engage from another router, it gains knowledge of the subnets and adds paths to its routing database. All routers in an internetwork can learn about certain subnets if they all participate. • Routing protocols also must avoid loops from happening when learning routes. A loop arises when a packet repeatedly returns to the very same router due to flaws in the routing tables of the collective routers. • If the routing protocol takes considerable effort to avoid loops, these loops can arise with routing protocols. • Routing protocols are mechanisms for exchanging routing information between routers and making routing decisions. • There are 3 kinds of routing protocols that are regularly used on the Internet. The distance vector, link state, and path vector are the three. 252 CU IDOL SELF LEARNING MATERIAL (SLM)

• The essential concepts and fundamentals underpinning every one of these three types of procedures are presented in this chapter in a general framework. Routing protocols, like distributed methods, are susceptible to traps such as looping during temporary periods. 14.7 KEYWORD  Hand-Off-The procedure by which a cellular phone call is passed from one radio frequency with one cell to another radio frequency in another cell by the Mobile Telephone Switching Office (MTSO).  Packet- For transmission, a collection of data structured in a specified way. The header, the content, and the trailer are the three main components of a packet (error detection and correction bits).  Spoofing- An access method in a switched network that provides a very quick dial-up routine and mimics the capabilities of a packet switched data network. 14.8 LEARNING ACTIVITY 1. Imagine you're a market manager, and you've got a new attractive product to offer. We are confident that the product will generate a large profit if it is sold to the proper people. So, from our vast customer base, how can we determine who is best suited for the product? ___________________________________________________________________________ ___________________________________________________________________________ 2. Take any kind of network topology. Determine the characteristics of them and set up a principle for clustering those networks. _______________________________________________________________________ _______________________________________________________________________ 14.9UNIT ENDQUESTIONS A. Descriptive Questions Short Questions 1. What are routing protocols? 2. How does routing algorithm helps in creating the routing table? 3. What are the metric values? 253 CU IDOL SELF LEARNING MATERIAL (SLM)

4. Write about adaptive algorithm. 5. What is centralized algorithm? Long Questions 1. Explain about classification routing algorithms. 2. Explain about non-adaptive algorithms. 3. Explain about flooding and random walk. 4. Describe about data dissemination. 5. Explain about clustering. B. Multiple Choice Questions 1. _____ protocols describe how routers communicate with one another a. Massive b. Network router c. Ransome model d. Unique 2. They _____ are routed using routing techniques a. router b. switches c. Mobility model d. packets 3. ____ is a measure that specifies the number of points a packet must make through internetworking devices. a. active model b. hop count c. metric value d. indoor 4. ______ of a link is the capacity of the link 254 a. Bandwidth CU IDOL SELF LEARNING MATERIAL (SLM)

b. MANET c. WAN d. Frequency 5. _____ algorithms that do not alter their routing decisions after they are chosen. a. adaptive b. non-adaptive c. neural d. WAN Answers 1-b, 2-d, 3- b, 4- a, 5-b 14.10REFERENCES Reference books • Behrouz A Forouzan, “Data Communications and Networking”, McGraw Hill. • Andrew S. Tanenbaum, “Computer Networks”, Pearson Education. • Subir Kumar Sarkar, T.G. Basavaraju, C. Puttaamadappa, “AdHoc Mobile Wireless Network: Principles, Protocols, and Applications, CRC Press. Textbook references • James F. Kurose, Keith W. Ross, “Computer Networking”, Pearson Education. • Michael A. Gallo, William M. Hancock, “Computer Communications and NetworkingTechnologies”, CENGAGE Learning Websites: • https://www.ciscopress.com/articles/article.asp?p=2180210&seqNum=7 • https://www.computernetworkingnotes.com/ • https://www.guru99.com 255 CU IDOL SELF LEARNING MATERIAL (SLM)

UNIT 15-SDN STRUCTURE 15.0 Learning Objectives 15.1 Introduction to SDN 15.2 Planes involved in SDN 15.3 SDN Architecture 15.4 Challenges of SDN 15.5 Advantages and Disadvantages of SDN 15.6 Applications of SDN 15.7 Summary 15.8 Keywords 15.9 Learning activity 15.10 Unit End Questions 15.11 References 15.0LEARNING OBJECTIVES After studying this unit, you will be able to • Learn the basics of SDN • Explain different planes involved in SDN • Describe the architecture of SDN • List the challenges of SDN • Outline the advantages and disadvantages of SDN 15.1INTRODUCTION TO SDN SDN (software-defined networking) is a networking architecture that abstracts the multiple, recognisable layers of a network to make it more agile and flexible. SDN aims to improve network control by allowing businesses and service providers to react swiftly to changing business needs. A network engineer or administrator may shape traffic in a software-defined network from a centralised control console without wanting to connect individual switches in the network. 256 CU IDOL SELF LEARNING MATERIAL (SLM)

Regardless of the actual connections between a server and devices, a centralised SDN controller will guide the switches to deliver internet services wherever they're needed. This technique different from previous network architecture, wherein individual network devices determine traffic decisions depending on the routing tables they have established. For almost a decade, SDN has played a role in networking, and its evolution & roles had continued to evolve. SDN (Software-Defined Networking) is a network architecture technique that allows networks to be intelligently & centrally controlled, or ‘programmed,' using software applications. Regardless of the specific network technology, this allows operators to manage and control network uniformly and holistically. SDN allows software applications to programme network behaviour in a remotely controlled manner utilising open APIs. Operators may manage the entire network as well as its devices uniformly regardless of the type of the supporting network technology by opening up formerly closed network platforms & creating a standard SDN control layer. 15.2 PLANES INVOLVED IN SDN In order to comprehend software defined networks, we must first comprehend the various networking planes. The data plane encompasses all actions involving and resulting on data packets sent from the end user. This includes the following: Packet transmission Data segmentation and re-assembly Packet replication for multicasting This plane contains all actions that are required to accomplish data plane operations but do not include end user data packets. In other terms, this is the network's brain. The control plane's responsibilities include: Tables for routing Creating policies for packet handling Each switch in a typical network does have its own data plane and control plane. The control planes of different switches share topological information and thus build a forwarding table that determines in which an arriving data packet should be routed via the data plane. The control plane is removed from the switch and assigned to a centralised unit termed the SDN controller in software defined networking (SDN). As a result, a network administrator may control traffic from a centralised panel rather than touching individual switches. The 257 CU IDOL SELF LEARNING MATERIAL (SLM)

data plane remains in the switch, and if a packet enters a switch, whose forwarding activity is determined by the entries in flow tables that the controller has pre-assigned. Match fields (such as input port number & packet header) and instructions make up a flow table. The packet is first compared to the flow table entries' match fields. The instructions for the associated flow entry are then carried out. The instructions could be to forward the packet through one or more ports, drop the packet, or add headers to it. The switch queries the controller, which sends a new flow item to the switch if a packet does not locate a corresponding match in the flow table. This flow entry determines whether the packet is forwarded or dropped by the switch. 15.3 SDN ARCHITECTURE • Three layers make up a typical SDN architecture. • Typical network applications such as intrusion detection, firewalls, and load balancing are found at the application layer. • The SDN controller, which serves as the network's brain, is part of the control layer. It also allows programmes created on top of it to have hardware abstraction. • Infrastructure layer: This layer is made up of hardware switches that compose the data plane and carry out actual data packet transfer. • The layers communicate using a set of interfaces known as northbound APIs (between the application and the control layer) and southbound APIs (between the application and the data layer) (between control and infrastructure layer). Fig 15.1 Layers of SDN 258 CU IDOL SELF LEARNING MATERIAL (SLM)

The application layer, the control layer, and the infrastructure layer are the three levels that make up a typical SDN architecture. The application layer, as expected, contains the common network applications and functions that businesses employ. Intrusion detection systems, load balancing, and firewalls are examples of this. A software-defined network replaces a typical network's specialised equipment, including a firewall or load balancer, with an application that employs a controller to regulate data plane behaviour. The SDN design divides the network in three distinct levels, which are connected via northbound & southbound APIs. The control layer is made up of the centralised SDN controller software that serves as the software-defined network's brain. This controller, which is housed on a server, is in charge of network policies and traffic flow. The physical switches throughout the network make up the infrastructure layer. The northbound & southbound programming languages are used to interact between these three layers (APIs). Applications, for example, communicate with the controller via its northbound interface, whereas the controller or switches communicate via southbound interfaces, like OpenFlow — although alternative protocols exist. There is no clear standard for both the controller's northbound API that match OpenFlow as a common southbound interface at the moment. Given its wide vendor support, the Open Daylight controller's northbound API is expected to become a de facto standard over time. How does SDN work? Functional separation, network virtualization, and automation through programmability are all examples of SDN technology. SDN technology was developed with the primary purpose of separating the network control plane and data plane. The data plane moves packets from one location to another while the control plane makes choices about how packets should flow through the network. 259 CU IDOL SELF LEARNING MATERIAL (SLM)

Fig 15.2 Architecture of SDN A packet arrives at a network switch in a basic SDN scenario, and rules incorporated into the switch's proprietary firmware inform the switch where to send the packet. The centralised controller sends these packet-handling rules to the switch. The switch, also called as a data plane device, asks the controller for direction when needed and gives information about traffic it processes to the controller. Every packet heading to same destination is sent along the same path by the switch, and all packets are treated the same way. A switch provides a route request to a controller for just a packet which does not have a specific path in software-defined networking, which is frequently referred to as adaptive or dynamic. Adaptive routing, on the other hand, sends route requests to routers and algorithms depending on the network topology, rather than a controller. A virtual overlay, which would be a logically independent network on top of the physical network, is where the virtualization part of SDN comes into play. End-to-end overlays can be used to hide the underlying network & segment network traffic. This micro segmentation is extremely important for service providers & operators with multi-tenant cloud environments or cloud services, as it allows them to create a different virtual network for each tenant, each with its own set of policies. Benefits of SDN SDN has a number of advantages, including: 260 CU IDOL SELF LEARNING MATERIAL (SLM)

With SDN, a network administrator can modify the rules of any network switch at any time, prioritising, deprioritizing, or even banning specific types of packets with granular control and security. This is particularly useful in a multi-tenant cloud computing architecture since it allows the administrator to control traffic loads in an efficient and versatile manner. Essentially, this allows the administrator to employ less expensive commodity switches while still having complete control over network traffic flow. Network management & end-to-end visibility are two more advantages of SDN. To deliver policies to connected switches, a network administrator only needs to deal with one centralised controller. This is in contrast to configuring numerous devices individually. So, because controller can control traffic and deploy security controls, this feature is also a security benefit. If the controller detects questionable traffic, for example, the packets can be rerouted or dropped. SDN also virtualizes technology and services that used to be performed by dedicated hardware. As a result, the claimed advantages of a smaller hardware footprint and lower operational expenses are realised. Software-defined networking also played a role in the development of software-defined wide area network (SD-WAN) technologies. The virtual overlay element of SDN technology is used in SD-WAN. This abstracts an organization's WAN connectivity lines, resulting in a virtual network that can transfer traffic over whatever connection the controller deems appropriate. SDN's Difficulties Service providers, network operators, telecommunications, and carriers are among the early adopters of SDN, as are huge organisations like Facebook and Google, which have the resources to handle and contribute to a developing technology. However, SDN is not without its difficulties. With SDN technology, security is both an advantage and a worry. The centralised SDN controller is a single point of failure that can be harmful to the network if it is targeted by an attacker. Another issue with SDN is that there is no universally accepted definition of \"software- defined networking\" with in networking industry. SDN is approached in a variety of ways by different vendors, ranging between hardware-centric models & virtualization platforms through hyper-converged networking and controllerless solutions. White box networking, network disaggregation, network automation, and programmable networking are all common networking approaches that are misunderstood for SDN. SDN is a different technology that may benefit from and work with existing technologies and procedures. 261 CU IDOL SELF LEARNING MATERIAL (SLM)

When the OpenFlow protocol and SDN technology were launched in 2011, there was a lot of buzz. Ever since, adoption has been gradual, particularly among smaller businesses with less resources and networks. The expense of implementing SDN is also a deterrent for many businesses. 15.4 CHALLENGES OF SDN New issues have evolved as a result of SDN. Because the fundamental functions of programmable switches have grown relatively independent of the hardware in use, the software component should provide effective switching capabilities. Both the control plane and the data plane required new methods or protocols (consider how the controller must configure the switches, for example). However, even with newer software tools, the data plane's features remain at a basic level in order to improve processing speed. However, this comes at a price in terms of new features and ease of use: adding a new feature (new protocol, changed network topology, etc.) may necessitate an upgrade of any and all data planes, posing a significant constraint in production environments. As a result, one of the problems is to design high-performance data planes while also exposing a powerful programmable, \"updatable\" interface. The importance of good software design cannot be overstated! However, hardware is not wholly disregarded. To get good performance, it is necessary to interface the short forms with the hardware devices in an effective manner. And getting good results is one of SDN's main goals! Bit rate performance, but also resource consumption—the more CPUs available for user applications, the better—or even other subsystems like storage: higher throughputs mean greater data, which must be routed to rapid and effective storage backends. Another major issue with SDN is security. The basic architecture gives birth to a decoupling of the control and data planes as the network topology evolves. This new architecture enables updating the network topology in real time considerably more possible and simpler. As a result, network components are more difficult to secure and monitor. It is critical, in particular, that directives on the control plane stay secure. And virtualization exacerbates the problem: when multiple appliances share a physical host, it must share all resources while avoiding leaking their data. 15.5 ADVANTAGES AND DISADVANTAGES OF SDN: • Because the network is programmable, it can be readily changed through the controller instead of individual switches. • Because each switch simply requires a data plane, switch hardware becomes more affordable. 262 CU IDOL SELF LEARNING MATERIAL (SLM)

• Because the controller is abstracted, applications can be created on top of it regardless of the switch vendor. • Because the controller could monitor traffic & deploy security controls, it provides superior security. The controller, for example, can reroute or discard packets if it detects suspicious behaviour in network traffic. Disadvantages: • SDN's disadvantages include a single point of failure due to the network's central reliance, which implies that if the controller becomes corrupted, the entire system would be affected. 15.6APPLICATIONS OF SDN Security Services are the first thing that comes to mind. Within the network layer, the modern virtualization environment offers unique virtual services. This entails introducing NFV capabilities into SDN platforms. This kind of network security enables a genuinely proactive environment that can reduce risk and respond to incidents much faster. When a security breach occurs, each second counts in halting the attack. The ability to detect the assault and guarantee that the other network resources are safe is also critical. We'll see more cyberattacks and much more sophisticated advanced threats as the network layer becomes much more crucial — and as the modern business becomes progressively more digitised. You can assist create a much more proactive environment that can respond to change by incorporating robust security agencies into the SDN layer. Network Intelligence & Monitoring is a term used to describe the process of gathering information from a Modern SDN technologies are assisting in the abstraction of one of the data centre’s most crucial layers: the network. Network infrastructures are far more complicated than ever before, and they must manage far more data. This means that knowing what's going on in your surroundings is crucial. Do you have troubles with latency on a certain port? What if you have network architecture that is heterogeneous? Or are you substantially virtualized, with a lot of traffic moving via the network layer? When you have a good network intelligence & monitoring layer, all of these issues go away. Integrating such technologies within your SDN architecture, on the other hand, provides true insight and benefits. Network intelligence and monitoring tools can incorporate traffic flow, port configurations, hypervisor integration, alerting, and even optimization. Most importantly, these types of flexible systems will aid in the monitoring of network traffic among your data centre and cloud ecosystem. Applications that are governed by regulations and compliance. The ability to store and interact with compliance-bound workloads is currently available from major cloud vendors. Organizations can now expand their architectures into distributed settings and the cloud, which were previously restricted by restrictions. However, how do you divide the traffic? 263 CU IDOL SELF LEARNING MATERIAL (SLM)

How do you keep compliance and regulation duties secure and under constant surveillance? This is when SDN comes in handy. In an SDN architecture, network traffic between switches, network points, or even hypervisors may all be controlled. This layer abstracts virtual operations and physical controls, so keep that in mind. This strong layer can then be applied to a variety of sites, virtualization points, and even cloud environments. Applications with a high level of performance. New types in application technologies are exploding. Rich apps like GIS, CAD, engineering, and graphics design software can now be delivered via virtualization. These workloads used to necessitate bare-metal architecture with their very own connections. Virtualization, on the other hand, allows apps to be streamed, and VDI can aid in the creation of sophisticated desktop experiences. However, we can see SDN being integrated into application control at the network layer. Creating robust QoS regulations, safeguarding secret data, segregating heavy traffic, or even setting threshold alarms around bottlenecks are all things that may be done. All of these SDN activities contribute to the delivery of high-performance, rich applications via virtualization. Cloud Integration and Distributed Application Control One of the most important features of SDN is its ability to span the whole data centre. Distributed locations, the cloud, and the entire business are all integrated in this form of agility. Critical network traffic can be routed between many sites using SDN, regardless of the underlying network architecture. You can transport data more easily between data centres and cloud locations by abstracting important network controls. Because SDN is a type of network virtualization, users can utilise strong APIs to control individual network services in addition to integrating with a cloud provider. This allows you to control your workloads more precisely while keeping your company adaptable. When you've a better understanding, keep in mind that additional SDN functionalities may have applications in your company as well. The trick, though, is to comprehend how SDN might benefit your data centre and your company. SDN fundamentally reduces the networking layer and allows you granular control over your dispersed data centre ecosystem's applications, services, and ecosystem. Most importantly, SDN assists you in creating a firm that can adapt to market upheavals and industry changes. This enables your company to be really flexible and effective. 15.7SUMMARY • As separating the control plane and data plane, software-defined networking provides various networking benefits. However, scalability, stability, and availability of networks continue to be major concerns. • Multi-controller architectures are therefore critical for SDN-enabled networks. 264 CU IDOL SELF LEARNING MATERIAL (SLM)

• This study provides an in-depth look of SDN multi-controller topologies. It introduces SDN and its primary implementation, OpenFlow. • The differences between various forms of multicontroller architectures, such as the distribution mechanism and the communication system, are then explained in detail. • It also describes the design, communication mechanism, and performance findings of multicontroller systems that have either been implemented or are currently being researched. • Multicontroller architectures in a software-defined network can have a variety of elements and characteristics, as we'll see in the following paragraphs, including the differences among logically or physically centralised and distributed architectures, as well as flat and hierarchical designs. 15.8 KEYWORDS • Time to Standby -The maximum amount of time that charging wireless portable and transportable phone can be used before the battery dies. Also see Talk Time. • Store-and-Forward -When the receiving device is unavailable, the ability to send a message to an intermediate network point and temporarily store it. • Synchronization-It is the act of transmitting data from previous or more databases such that each is identical. It is also known as \"replication.\" 15.9 LEARNING ACTIVITY 1. Draw the architecture of SDN for the network which is set in your organization. ___________________________________________________________________________ _________________________________________________________________ 2. Analyse the challenges of SDN and try to find solutions for the same ___________________________________________________________________________ _____________________________________________________________ 15.10 UNIT ENDQUESTIONS A. Descriptive Questions Short Questions 1. What is SDN? 2. What is control layer? 3. What are the benefits of SDN? 265 CU IDOL SELF LEARNING MATERIAL (SLM)

4. List the difficulties of SDN. 5. What is updatable interface? Long Questions 1. Explain the planes involved in SDN. 2. Explain the architecture of SDN. 3. How does SDN works? 4. Describe the challenges of SDN. 5. Explain the issues with SDN security. B. Multiple Choice Questions 1. ____ is a network architecture technique that allows networks to be intelligently & centrally controlled. a. Program b. SDN c. Ransome model d. Topology 2. The layers communicate using a set of interfaces known as northbound _____ a. APIs b. switches c. data planes d. packets 3. Which of the following is not an example of SDN? a. network virtualization b. functional separation c. network equipment d. automation 4. The switch, also called as a ___ device 266 CU IDOL SELF LEARNING MATERIAL (SLM)

a. Bandwidth b. mobility c. data plane d. control 5. OpenFlow protocol is launched in ____ a. 2000 b. 2011 c. 2010 d. 2014 Answers 1-b, 2- a, 3- c, 4- c, 5-b 15.11REFERENCES Reference books • Behrouz A Forouzan, “Data Communications and Networking”, McGraw Hill. • Andrew S. Tanenbaum, “Computer Networks”, Pearson Education. • Subir Kumar Sarkar, T.G. Basavaraju, C. Puttaamadappa, “AdHoc Mobile Wireless Network: Principles, Protocols, and Applications, CRC Press. Textbook references • James F. Kurose, Keith W. Ross, “Computer Networking”, Pearson Education. • Michael A. Gallo, William M. Hancock, “Computer Communications and NetworkingTechnologies”, CENGAGE Learning Websites: • https://www.ciena.com/insights/what-is/What-Is-SDN.html • https://www.computernetworkingnotes.com/ • https://opennetworking.org/sdn-definition/ 267 CU IDOL SELF LEARNING MATERIAL (SLM)