Computing as well as networking systems usually require load balancing frameworks. P2P networks have a large number of proposed load balancing mechanisms. They achieve this by improving the availability limit in P2P systems. Well their convergence speed is still quite low and they keep a high load transfer overhead as well as load collection overhead. Our system proposes an improved load balancing system based on Ant based load balancing algorithm. It consists of target nodes selection framework along with candidate node discovery system. Here 2 categories of agents coordinate in order to achieve this (the Search and the Guide ant). In order to get a desired load balancing effect as compared to classic algorithms, factors such as load collection overhead as well as load transfer overhead of our algorithm are lowered gradually, this is validated through a performance evaluation. Our system shows the effect algorithm parameters have on performance of the proposed algorithm.