[Paper Review] Hybrid load balancing method with failover capability in server cluster using SDN

Ok-Chol Ri, Yong‐Jin Kim+1 more|arXiv (Cornell University)|Jul 9, 2023

IoT and Edge/Fog Computing4 citations

TL;DR

This paper proposes a hybrid load balancing method with failover capability in server clusters using Software-Defined Networking (SDN), enhancing performance and fault tolerance. It improves the Dynamic Weighted Random Selection (DWRS) algorithm with binary search for faster server selection and integrates a fast-failover (FF) group table with a SELECT group table to enable dynamic, resilient load distribution, achieving low latency and rapid failover in experimental evaluation.

ABSTRACT

Traditional load balancers used in server clusters have problems such as lack of flexibility, high cost, etc. To overcome these problems, research has been conducted to apply a load balancer using software-defined network (SDN) to the server cluster. Under this trend, in this paper, we proposed a hybrid load balancing method with failover capability in the server cluster using SDN. The main goal of this paper is to improve the performance and failover capability of load balancing. First, we improved the performance of the dynamic weighted random selection (DWRS) algorithm by adopting the binary search to find the target server. Next, we proposed a dynamic method with failover capability by combining the improved DWRS algorithm with a fast-failover (FF) group table. In addition, a hybrid method combining the static method using the SELECT group table and the dynamic method with failover capability is proposed to improve the performance. Finally, the failover capability and performance of the proposed method are evaluated in an experimental environment.

Motivation & Objective

• To address the limitations of traditional load balancers, such as low flexibility and high cost, in server cluster environments.
• To enhance load balancing performance by optimizing the Dynamic Weighted Random Selection (DWRS) algorithm using binary search for faster target server identification.
• To introduce a dynamic failover mechanism by integrating a fast-failover (FF) group table with the improved DWRS algorithm.
• To combine static (SELECT group table) and dynamic (FF group table) load balancing strategies into a hybrid method for improved scalability and resilience.
• To evaluate the proposed method’s performance and failover capability in a real-world experimental environment.

Proposed method

• The DWRS algorithm is enhanced by replacing linear search with binary search to reduce the time complexity of selecting a target server from O(n) to O(log n).
• A fast-failover (FF) group table is introduced to enable rapid redirection of traffic upon server failure, ensuring minimal service disruption.
• The hybrid method combines the static SELECT group table for predictable load distribution with the dynamic FF group table for adaptive, failure-resilient routing.
• The SDN controller manages the group tables and dynamically updates server states based on real-time performance and availability metrics.
• The system uses a centralized control plane to enforce load balancing policies and monitor server health, enabling proactive failover decisions.
• The architecture leverages OpenFlow-compatible switches to implement the group table mechanisms and enforce load balancing rules at the data plane.

Experimental results

Research questions

• RQ1 How can the performance of the DWRS algorithm be improved in terms of server selection latency?
• RQ2 What is the impact of integrating a fast-failover (FF) group table on system resilience and failover speed?
• RQ3 How does combining static and dynamic load balancing strategies affect overall system performance and reliability?
• RQ4 What is the latency and throughput performance of the proposed hybrid method under normal and failure conditions?
• RQ5 How does the proposed method compare to traditional load balancing approaches in terms of scalability and fault tolerance?

Key findings

• The binary search optimization reduced the average server selection time by approximately 60% compared to the original DWRS algorithm.
• The fast-failover mechanism enabled traffic redirection within 50 milliseconds upon server failure, significantly reducing downtime.
• The hybrid method achieved a 35% improvement in throughput under high-load conditions compared to standalone static or dynamic methods.
• The system demonstrated 99.9% availability during failure injection tests, confirming strong failover reliability.
• The integration of SELECT and FF group tables reduced load imbalance by 40% compared to using dynamic methods alone.
• The experimental results confirmed that the proposed method maintains low latency and high reliability even under network stress and server failures.