To improve the efficiency of resource access in Software as a Service (SaaS) applications, a load balancing strategy combined with the important features of SaaS service was proposed. Firstly, the load balancing strategy was proposed by combining multi-tenancy and high scalability in SaaS service based on the distribution of request and global and local scalability. Secondly, the load balancing strategy model was constructed and simulated by a Petri net. Finally, this strategy was compared with Round Robin (RR), stochastic algorithm and Improved Least-Connection Scheduling (ILCS) load balancing algorithm in response time and throughput. The experimental results show that the response time and throughput of the proposed strategy become stable and they are superior to the other three algorithms after the request rate reaches 500 per second.