SDN Controller Placement With Availability Upgrade Under Delay and Geodiversity Constraints

Abstract

An inherent problem in Software-Defined Networking (SDN) is the Controller Placement Problem, which addresses how many controllers to deploy in the network, and where to place them. Several variants of this problem have been addressed and researched to find the placements that adapt best to different contexts. In this paper, we address a more complex variant of this problem, to satisfy QoS requirements and to offer robustness against disaster-based failures. We address the joint optimization problem of controller placement and finding a tree subgraph which can be upgraded to have enhanced availability, in order to satisfy delay and availability constraints. Additionally, we consider geodiversity constraints as a way to enhance robustness to disaster-based failures.

Joint optimization of primary and backup controller placement and availability link upgrade in SDN networks

“…These results are compared with those in [18], to compare the gains in terms of upgrade costs of having either path redundancy to the primary controller, or having path redundancy to a primary and a backup controller (controller redundancy). A brief analysis between path versus controller redundancy is also present in [19], where we focused mainly on the performance of the geodiverse solutions. Finally, comparison between the solutions and runtimes of the joint optimization ILP model with the heuristic method described in Section 6, are discussed.…”

“…Constraints (18) guarantee that variables 0 are assigned to the non-upgraded arcs of the backup path from node to its backup controller. Constraints (19) guarantee that variables are assigned to the arcs of the backup path for node to its backup controller, that are upgraded to level = 1, ..., . Constraints (20) guarantee that each link is upgraded to one and only one specific level , or is not upgraded at all.…”

“…We further extended the work in [18] to include path redundancy and considered the spine to be a spanning tree. In [19], we considered this framework where geodiversity guarantees were further imposed.…”

Cost-Effective Survivable Controller Placement in Software-Defined Networks

“…Zhang et al [17] investigated different topologies with considering different locations of the controllers. Santos et al [18] used a tree sub graph to optimize the controller, which can be updated to increase availability. Furthermore, some limitations on latency and availability were considered in the mathematical model of the problem.…”

“…E itself contains the links connecting the controller to the switch and the controller to the controller indicated by the symbols E F and E S , respectively. Links are considered directional when the process of controller placement and Methods Objectives Latency Scalability Reliability Cost [7] The average latency and worst-case latency * --- [10] Integer linear programming * --* [11] Integer linear programming * -* - [12] The spectral clustering algorithm -* * - [13] Determining the failure rate --* - [14] Graph parsing algorithm * -* - [15] The spanning tree topology * --* [16] Using an efficient algorithm * --- [17] Cluster of multiple controllers * * * - [18] Tree subgraph * -* - [19] Hierarchical clustering * * * - [20] Integer linear programming * * -* [21] Using the clique algorithm * * * - [22] Varna optimization method * * * - [23] Bee colony algorithm * * -* [24] RetroFlow algorithm * -* - [25] Programmability guardian * * * - [26] Integer linear programming * -* - [27] Matchmaker assigning switches to them is called. In this way, the switches are connected to their controllers and the controllers are connected to each other.…”