As web services and applications become increasingly complex and user demand grows—especially with the rising number of students—the need for a reliable E-Learning system becomes critical. At Politeknik Negeri Balikpapan, the current E-Learning platform operates on a single server, leading to slow response times and potential server downtime under high traffic conditions. This study addresses the issue by implementing load balancing using two algorithms: Round Robin and Least Connection, across three web servers and one separate database server. Testing was conducted using Apache JMeter with 1000 requests per 10 seconds. Results show that the Least Connection algorithm outperformed Round Robin, achieving an average response time of 155.8ms, compared to 184.2ms. Compared to the single-server setup, the load-balanced system showed significant improvements in response time, error rate, concurrency, availability, upstream, and downstream metrics. CPU load was also reduced due to traffic distribution across multiple servers. This demonstrates that server resource optimization via load balancing can significantly enhance the overall performance of E-Learning services. These findings provide a strong foundation for more efficient and scalable IT infrastructure development and support better decision-making in managing high-demand educational platforms in the future