Load Balancing Method Performance Analysis on Haproxy and Router OS

Abstract

Increasing requests for client connections to servers on a network can cause a problem with server performance to decrease and too many user requests make the workload on the server increase rapidly resulting in the server being down, so we need technology that handles complex network usage. Then it can be solved with a load balancing system. Several studies on load balancing on haproxy and router OS with various algorithms applied will produce different performance, including round robin, least connections, and source algorithms on haproxy load balancing, while NTH, PCC, and ECMP on load balancing routers OS. Testing is done with 3 scenarios, scenario 1 with a load of 10 users, scenario 2 with a load of 20 users, and scenario 3 with a load of 30 users. The test parameters used are throughput, delay, and jitter using the webserver stress tool and wireshark. The results of the performance measurement of the load balancing method on haproxy generally show that the least connection is better than the round robin and source methods, this can be seen from the measurement figures for delay and jitter parameters that are best found in the loast connection method. While the results of the performance measurement of the load balancing method on the OS router, there are two, one shows that PCC is better, this can be seen from the measurement numbers of the delay and jitter parameters, then the second shows that the ECMP method is better, this can be seen from the throughput parameter measurement figures and delay.