Layer-4 Load Balancer on Programmable Data Plane using IP Hash and Weighted Round Robin

Layer-4 load balancer used in data center to distribute requests coming from client to multiple servers located in data center. There are two requirements to build proper Layer-4 load balancing. The first is ensuring the uniformity of request distribution across multiple servers and connection affinity. IP Hash is a common load balancing algorithm that is used to implement Layer-4 load balancing but can cause load imbalances across multiple servers. Weighted Round-Robin is proposed to prevent load imbalance across multiple servers. The implementation of load balacing is using a programmable data plane rather than additional hardware or software load balancing. The result shows that WRR is able to mitigate load imblance across multiple servers by achieving uniform request distribution. WRR achieves 13% higher throughput but gives 2% higher response times than IP Hash and has less packet loss than IP Hash when handling HTTP requests.

Keywords: keyword Layer-4, Load Balancer, Connection Affinity, IP Hash, Weighted Round Robin, Data Plane

Authors who publish in this journal agree to the following rules:

1. Authors retain copyright and give the journal the right of first publication, with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
2. Authors may enter separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., posting it to an institutional repository or publishing it in a book), with attribution to the journal's initial publication.
3. Authors are permitted and recommended to post their work online (such as in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges as well as earlier and greater citation of published work.