Thisarticle discusses the best in class in traffic engineering for SDN withattention to four cores including flow management, fault tolerance, topologyupdate, and traffic analysis. The challenging issues for SDN trafficengineering solutions are talked about in detail.
SDN is a developingnetworking administration that isolates the network control plane from the datasending plane with the guarantee to significantly enhance organize asset usage,streamline network management, reduce working expenses, and promote developmentand advancement. For network performance optimization by dynamically analyzing,predicting, and regulating the behavior of the transmitted data, trafficengineering is an important subject. Over the last two decades trafficengineering has been exploited in asynchronous transfer mode(ATM) andIP/multiprotocol label switching(MPLS). The existing traffic engineeringtechnologies rely on closed and firm architectural design, where the controland data planes are firmly coupled and coordinated. The inflexibility and theclosed nature of the architecture prevents the existing traffic engineeringtechnologies to provide differentiated services that would adapt toincreasingly growing uneven and highly variable traffic patterns.
Coming backto the four cores flow management, fault tolerance, topology update, andtraffic analysis: Starting of with the flow management, in SDN the significantoverhead at both control and data planes caused when an OF switch receives anew flow that doesn’t match any rule in its flow entry. To overcome the bottleneck in SDN solution are discussed in the paper, by considering the balancebetween latency and load balancing at the data plane and control plane. In loadbalancing for the data plane, two methods are followed: Hash-Based ECMP FlowForwarding and Wildcard Rule Flow Forwarding. The hash-based Equal-CostMulti-Path(ECMP) is a load balancing scheme to distribute flows acrossavailable paths using flow hashing methods.
This solution can improve the bandwidth and processing overhead at the switches. In wildcard rule flow forwardingto aggregate traffic flows, OF switches use flow match wild cards. In loadbalancing for the control plane SDN may cause the network controller to becomea potential performance bottleneck, to overcome this controller load balancingschemes have been used. They are logically centralized and physicallydistributed controller, physically distributed controller, hierarchicalcontroller, hybrid controller, multithread controllers and generalizedcontrollers. The multithread controllers are developed to improve the requestprocessing throughput. The parallelism architecture of servers is used toprovide high throughput with scalability at the controllers in multithreadmulticore SDN controllers.
The research challenges in flow management are dynamicload balancing scheme for the data plane and dynamic load balancing scheme forthe control plane. Now coming to the second core i.e. fault tolerance: forensuring the reliability of the network, SDN must be able to perform thefailure recovery straight forwardly and effortlessly when a failure occurs inthe network. Even though a switch could be used in such a failure, it hasneither the insight nor the information to make new route. For this it has todepend on the controller. In this manner, we discuss the research efforts onfast failure recovery in SDN networks. For fault tolerance in the data plane wehave two failure recovery mechanisms, which are restoration and protection.
In restoration,the recovery path can be pre-planned or dynamically designated. Whereas inprotection the recovery path must be pre-planned and reserved before thefailure occurs. Taking the bandwidth and latency factors into consideration,for larger SDN systems the protection solution is used for faster failurerecovery. Inthe control plane the fault tolerance is an important factor to maintain theefficiency of the network. A single point failure can cause an entire networkto degrade. The primary-backup replication method is the most fundamentalmethod used in the recovery process of the control plane.
The OF protocol donot have any coordination mechanisms for the primary and backup controllers touse. For this reason, the coordination protocols are wanted. In the deploymentof the backup controller, properly placing the backup controllers in the SDN wecan increase the network reliability. There are mainly two research challenges in thefault tolerance, they are cost-efficient and fast failure recovery for the dataplane and primary-backup replication with the traffic adaptivity for thecontrol plane. Now coming to the third core i.e. topology update. In thetopology update pre-planned changes like network policy rules change.
Whenthese changes occur, each flow is identified and then it is guaranteed to bemanaged by either the old or the new policy. There are two types of consistency:per-packet consistency and per-flow consistency, which have a common updateoperation that is the switches process the packet by following the old or newpolicies until the old configuration rules are deleted by the controller. Forthe problem which is caused by the duplicate policy scheme we require more efficientupdate algorithm for implementing consistent update.
The research challengesthat arises during the topology update are a single controller in a large scaleSDN network and multiple controllers in multi-domain SDN networks. Now talkingabout the fourth or the last core i.e. traffic analysis. In traffic analysis monitoringthe network is significant for the network management.
When it comes to SDNnetworks, the existing monitoring solutions become less efficient due to thelarge-scale nature of the network which effects the central controller.Query-base and Push-based monitoring are the current solutions available. TheQuery-based solution is based on the request or response paradigm, whereas thePush-based solution is based on publish or subscribe or distribute paradigm. The research challengesthat come up during the traffic analysis are traffic analysis, traffic monitoringand network checking and programming debugging methods.
In this paper, theinvestigation is done on the SDN traffic engineering solutions form the variousaspects of flow management, load balancing, fault tolerance, topology updateand traffic analysis. The current state and research difficulties of SDNtraffic engineering are exhibited by the addressing the key SDN executionmeasurements as far as scalability, availability, reliability, consistency andaccuracy.