Topology-Aware Controller Assignment in SDN Clusters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Software-defined networks (SDNs) face challenges in optimizing controller assignments and minimizing downtime due to the lack of consideration for physical topology and traffic implications, leading to suboptimal uptime and performance.
Innovation Solution
The technology configures SDNs by organizing nodes into clusters based on physical topology, assigning each cluster to a controller, and using graph-partition algorithms to minimize traffic between clusters, ensuring high availability and efficient traffic management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If nodes are assigned to controllers without considering physical topology, then controller assignment is simple, but network downtime increases and resilience decreases
Solution Approach 1:
The patent segments the SDN network into multiple clusters based on physical topology, with each cluster assigned to a specific controller. This segmentation isolates controller failures to specific clusters, preventing cascade failures across the entire network and thereby improving resilience without requiring complex global coordination mechanisms.
Solution Approach 2:
The patent performs preliminary clustering of nodes based on physical topology before controller assignment. By pre-organizing nodes into topology-based clusters and assigning controllers to these clusters in advance, the system prepares for potential failures before they occur, reducing downtime when failures happen without requiring complex real-time decision-making.
2Productivity
If controllers manage all nodes without topology consideration, then controller management is simplified, but traffic optimization deteriorates
Solution Approach 1:
The patent applies local quality by configuring each cluster with specific traffic optimization characteristics based on its physical topology. Each cluster can have tailored traffic engineering policies and optimizations suited to its local network structure, improving overall traffic efficiency without requiring a single complex global configuration.
Solution Approach 2:
By segmenting the network into topology-based clusters, the patent enables independent traffic optimization within each cluster. This segmentation allows traffic efficiency improvements in each local cluster without affecting other clusters, avoiding the need for complex global traffic management while still achieving overall optimization.
3Loss of time
If controller failures affect multiple tree structures, then system complexity is reduced, but downtime increases
Solution Approach 1:
The patent segments the network into isolated clusters where each cluster contains a specific tree structure and its associated controller. This segmentation ensures that controller failures are contained within single clusters, preventing cascade failures across multiple tree structures and reducing overall downtime without requiring complex isolation mechanisms.
Solution Approach 2:
The patent introduces cluster boundaries as intermediaries that isolate controllers and their associated tree structures. These intermediary cluster boundaries prevent failure propagation between controllers while maintaining manageable system complexity, as the isolation is achieved through simple cluster delimiters rather than complex inter-controller coordination mechanisms.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Control plane management based on physical topology of nodes in software-defined networks (SDNs) is provided. Nodes in an SDN are organized into a plurality of node clusters. A cluster of nodes in an SDN is formed and control plane management for the nodes within the cluster is provided. A topology-aware assignment of nodes to controllers in an SDN is provided. Related nodes such as root and corresponding child nodes are assigned to a single cluster in one embodiment to lessen the impacts of controller failure. The assignment of nodes to controllers is also provided based on minimizing traffic between node clusters and between nodes and external networks.