Limits...
Distributed controller clustering in software defined networks

View Article: PubMed Central - PubMed

ABSTRACT

Software Defined Networking (SDN) is an emerging promising paradigm for network management because of its centralized network intelligence. However, the centralized control architecture of the software-defined networks (SDNs) brings novel challenges of reliability, scalability, fault tolerance and interoperability. In this paper, we proposed a novel clustered distributed controller architecture in the real setting of SDNs. The distributed cluster implementation comprises of multiple popular SDN controllers. The proposed mechanism is evaluated using a real world network topology running on top of an emulated SDN environment. The result shows that the proposed distributed controller clustering mechanism is able to significantly reduce the average latency from 8.1% to 1.6%, the packet loss from 5.22% to 4.15%, compared to distributed controller without clustering running on HP Virtual Application Network (VAN) SDN and Open Network Operating System (ONOS) controllers respectively. Moreover, proposed method also shows reasonable CPU utilization results. Furthermore, the proposed mechanism makes possible to handle unexpected load fluctuations while maintaining a continuous network operation, even when there is a controller failure. The paper is a potential contribution stepping towards addressing the issues of reliability, scalability, fault tolerance, and inter-operability.

No MeSH data available.


Illustrated controller placement options.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC5383233&req=5

pone.0174715.g002: Illustrated controller placement options.

Mentions: Fig 2 shows different controller placement scenarios using one or two controllers with five switches. In scenario 1, a single controller controls the switches and in the other scenarios with two controllers each, either switch can be connected to any of the controllers. The first scenario with one controller connecting the five switches is less reliable than the other scenarios because a single controller presents the single point of failure problem. The use of more than one controller also affects the reliability of the network, for instance, in the second and third scenarios where both are with two controllers but placed differently, the third scenario is more reliable than the second scenario; because when the link between switch A and switch B fails in the second scenario, the communication path between switch A and its controller is broken but when there is any link failure in the third scenario, there will be at least one communication path available, which makes it more reliable than the second scenario. The way the switches are connected to the controller also affects the reliability of the network because, in the fourth scenario, the switches are placed in exactly the same locations as the third scenario but this time switch A is controlled by controller 1 in the fourth scenario instead of controller 2. Therefore, this makes the fourth scenario less reliable than the third scenario because when there is a link failure between switch B and switch C, the communication link from both switch A and switch B to their controller which is controller 1 will be broken. As regards this observation, the controllers in the proposed system are placed using the capacitated controller placement algorithm CCPP. The (CCPP) considers the load of the controllers for its placement algorithm. The aim is to reduce the number of required controllers and analyze the load of controllers, which is mainly processing packetIn events and delivering the events to the applications. The CCPP was defined as a variant of the k-center problem and has two phases. In phase one, the lower bound of radius is obtained in binary search and in phase two, the radius is increased from lower bound until a placement is found The algorithm considers only possible distances in phase one i.e. the distance between any pair of locations rather than all integers in a given range that ensures a faster convergence. The binary search converges until the step is less than 1; it requires more iteration than searching in possible distances because the possible radius must be one of the distances, searching in possible distances will not omit the result radius. This ensures that the algorithm always finds the exact location.


Distributed controller clustering in software defined networks
Illustrated controller placement options.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC5383233&req=5

pone.0174715.g002: Illustrated controller placement options.
Mentions: Fig 2 shows different controller placement scenarios using one or two controllers with five switches. In scenario 1, a single controller controls the switches and in the other scenarios with two controllers each, either switch can be connected to any of the controllers. The first scenario with one controller connecting the five switches is less reliable than the other scenarios because a single controller presents the single point of failure problem. The use of more than one controller also affects the reliability of the network, for instance, in the second and third scenarios where both are with two controllers but placed differently, the third scenario is more reliable than the second scenario; because when the link between switch A and switch B fails in the second scenario, the communication path between switch A and its controller is broken but when there is any link failure in the third scenario, there will be at least one communication path available, which makes it more reliable than the second scenario. The way the switches are connected to the controller also affects the reliability of the network because, in the fourth scenario, the switches are placed in exactly the same locations as the third scenario but this time switch A is controlled by controller 1 in the fourth scenario instead of controller 2. Therefore, this makes the fourth scenario less reliable than the third scenario because when there is a link failure between switch B and switch C, the communication link from both switch A and switch B to their controller which is controller 1 will be broken. As regards this observation, the controllers in the proposed system are placed using the capacitated controller placement algorithm CCPP. The (CCPP) considers the load of the controllers for its placement algorithm. The aim is to reduce the number of required controllers and analyze the load of controllers, which is mainly processing packetIn events and delivering the events to the applications. The CCPP was defined as a variant of the k-center problem and has two phases. In phase one, the lower bound of radius is obtained in binary search and in phase two, the radius is increased from lower bound until a placement is found The algorithm considers only possible distances in phase one i.e. the distance between any pair of locations rather than all integers in a given range that ensures a faster convergence. The binary search converges until the step is less than 1; it requires more iteration than searching in possible distances because the possible radius must be one of the distances, searching in possible distances will not omit the result radius. This ensures that the algorithm always finds the exact location.

View Article: PubMed Central - PubMed

ABSTRACT

Software Defined Networking (SDN) is an emerging promising paradigm for network management because of its centralized network intelligence. However, the centralized control architecture of the software-defined networks (SDNs) brings novel challenges of reliability, scalability, fault tolerance and interoperability. In this paper, we proposed a novel clustered distributed controller architecture in the real setting of SDNs. The distributed cluster implementation comprises of multiple popular SDN controllers. The proposed mechanism is evaluated using a real world network topology running on top of an emulated SDN environment. The result shows that the proposed distributed controller clustering mechanism is able to significantly reduce the average latency from 8.1% to 1.6%, the packet loss from 5.22% to 4.15%, compared to distributed controller without clustering running on HP Virtual Application Network (VAN) SDN and Open Network Operating System (ONOS) controllers respectively. Moreover, proposed method also shows reasonable CPU utilization results. Furthermore, the proposed mechanism makes possible to handle unexpected load fluctuations while maintaining a continuous network operation, even when there is a controller failure. The paper is a potential contribution stepping towards addressing the issues of reliability, scalability, fault tolerance, and inter-operability.

No MeSH data available.