An SDN-Based Solution for Horizontal Auto-Scaling and Load Balancing of Transparent VNF Clusters

Area of Science:

• Computer Science
• Network Engineering
• Distributed Systems

Background:

• Virtualized network functions (VNFs) present challenges in dynamic scaling and load balancing.
• Maintaining bidirectional flow affinity and avoiding loops during scaling-out are critical issues.

Purpose of the Study:

• To propose and validate a software-defined networking (SDN)-based solution for dynamic scaling and load balancing of VNFs.
• To address complexities like loop avoidance and bidirectional flow affinity in VNF scaling.

Main Methods:

• Implementation of an SDN-based solution using two SDN controllers and two OpenFlow switches (OFSs).
• SDN controllers manage monitoring, scaling, and load-balancing logic.
• OFSs redirect traffic based on controller instructions, employing a CPU-based load-balancing strategy.

Main Results:

• Successful end-to-end (E2E) traffic connectivity through VNF clusters was achieved.
• Bidirectional flow affinity strategy performed well, creating simultaneous flow rules.
• CPU-based load balancing ensured average imbalance below 10% without packet modification.
• Failure detection occurred in near real-time, with new instance activation under a minute.
• Auto-scaling module responded quickly to traffic fluctuations.

Conclusions:

• The proposed SDN and OFS solution offers high flexibility for implementing diverse load-balancing, scaling, and monitoring strategies.
• The system effectively handles dynamic scaling and load balancing for VNFs.
• The solution demonstrates practical feasibility on a real testbed.