An Effective Approach to Controller Placement in Software Defined Wide Area Networks

Abstract

Abstract-One grand challenge in Software Defined Networking (SDN) is to select appropriate locations for controllers to shorten the latency between controllers and switches in wide area networks. In the literature, the majority of approaches are focused on the reduction of packet propagation latency, but propagation latency is only one of the contributors of the overall latency between controllers and their associated switches. In this paper, we explore and investigate more possible contributors of the latency, including the end-to-end latency and the queuing latency of controllers. In order to decrease the end-to-end latency, the concept of network partition is introduced and a Clustering-based Network Partition Algorithm (CNPA) is then proposed to partition the network. The CNPA can guarantee that each partition is able to shorten the maximum end-to-end latency between controllers and switches. To further decrease the queuing latency of controllers, appropriate multiple controllers are then placed in the subnetworks. Extensive simulations are conducted under two real network topologies from the Internet Topology Zoo. The results verify that the proposed algorithm can remarkably reduce the maximum latency between controllers and their associated switches.

Effective Scheduling Function Design in SDN Through Deep Reinforcement Learning

“…BLAC [10] randomly sampled a small number of controllers and sent requests to the least loaded one. Similar approaches can also be found in literature [8], [22]. Although such manually designed SFs are intuitive and simple in nature, the design process is time-consuming and requires substantial domain knowledge.…”

Toward a Flexible Design of SDN Dynamic Control Plane: An Online Optimization Approach

“…Dyn. Main Optimization Metrics(s) [14] Average control latency [11] Average flow setup time [31] Control plane operational cost [15] Control latency & load imbalance [32] Controller utilization and operational cost [16] Average control latency [17] Max. control latency upon controller failures [18] Avg.…”

“…Therefore, we can model each controller as an M/M/1 queue. This model has been validated to approximate the processing time of a controller with one switch [39] and multiple switches [40], and it has been applied in previous work of CPP for SD-WAN [16]. Nevertheless, other more accurate queuing models can be incorporated and therefore extend our problem formulation.…”

“…In order to be able to solve the offline optimization problem with linear optimizer, such as Gurobi and CPLEX, we need to linearize the non-linear equations in Problem P. The high order constraints, i.e., Constraints (11), (13), (15), (16) and (17), can be linearized without loss of optimality [11]). For the definition (18), however, we need to apply piecewise linear approximation and define the following linear relationship between μ (t) Φ (φ) (the expected processing time of a flow setup request at a controller) and λ (t) Φ (φ) (the total number of flow setup requests the controller needs to process),…”

“…This is an algorithm we have adopted from our previous work [11] which does not consider control plane reconfiguration. (2) CNPA: We take another state-of-theart algorithm adapted from [16] which is designed based on topology clustering. We need to amend the original algorithm, so that the clustering can take the dynamicity of the flows into account.…”