Scalability of a Desktop Grid project is ultimately constrained by the server’s maximum request processing capacity, which can be extended through hardware upgrades or load balancing but remains a fundamental limitation. This capacity is influenced by three key factors: the total number of active nodes, the processing complexity of each node interaction, and the average computation time required per task. In this paper a mathematical model aimed to BOINC performance optimization is presented. The model represents a Desktop Grid system as a queueing system M/M/1/m having one server and buffer of size m. The model could be used to identify the optimal size of a task to maximize the server performance.