Tip decomposition has a pivotal role in mining cohesive subgraph in bipartite graphs. It is a popular research topic with wide applications in document clustering, spam group detection, and analysis of affiliation networks. With the explosive growth of bipartite graph data scale in these scenarios, it is necessary to use distributed method to realize its effective storage. For this reason, this work studies the problem of tip decomposition on a bipartite graph in the distributed environment for the first time. Firstly, a new relay based communication mode is proposed to realize effective message transmission when the given bipartite graph is decomposed in distributed environment. Secondly, the distributed butterfly counting algorithm (DBC) and tip decomposition algorithm (DTD) are designed. In particular, a controllable parallel vertex activation strategy is proposed to solve the problem of memory overflow when DBC decomposes large-scale bipartite graphs. Finally, the message pruning strategy based on vertex priority and message validity pruning strategy are introduced to further improve the efficiency of the algorithm by reducing redundant communication and computing overhead. The experiment is deployed on the high performance distributed computing platform of National Supercomputing Center. The effectiveness and efficiency of the proposed algorithms are verified by experiments on several real datasets.