Smart machines achieve daily life in-demand services increasingly, and quicker access to required services is of great importance and, in some services, is very critical. In this paper, a new architecture is proposed to improve the distributed cache proxy server performance in distributed systems for demanded services by developing cookie-session middleware and employing polymorphism mechanism in DHT servers. By using node clustering, each node as a peer is a server/client that listens to the line on the TCP port and uses session-cookie middleware for the requested service. When it is only a node, it is a server/client that listens to the line on the UDP port and performs DHT protocol for registration of the peers’ characteristics. An architecture based on the game theory has been proposed in this research. Moreover, an extensive-form game with imperfect information, called weak sequential equilibrium, has been studied and proven. It has weak sequential and Nash equilibriums in the state of ((mtcp, mudp), (Y, Y); λ =1, μ=0)). The requests prioritization in the cookies has been studied, and load balancing of the sessions has been performed based on Round Robin prioritization. The simulation results show that the new architecture is optimal in terms of bandwidth consumption, user response time, and cache memory hit ratio.