The invention provides a server load balancing method and device. The method includes receiving and analyzing user-submitted application requests, application identifiers belonging to the application requests, and application request information of a deployed server list of applications, and arranging at least one application in each server in the server list; transmitting the application requests to the servers with the throughput smaller than the throughput threshold in the server list to process. The throughput of each server and a throughput threshold can be acquired according to the request arrival rate of each application arranged in each server and the request arrival rate threshold in the single application environment. According to the server load balancing method and device, the application requests are transmitted to the servers with the throughput smaller than the throughput threshold, and service quality requirements of all applications in the servers are guaranteed effectively.