Cluster ensembles are approaches to combine different clustering results to obtain a robust consensus partitioning. However, many cluster ensemble methods suffer from the problem of scalability since the extensive cost of calculating co-association matrix, which makes it hard to perform cluster ensemble on large scale datasets. In this paper, we proposed a scalable co-association cluster ensemble framework using a compressed version of co-association matrix formed by selecting representative points of origin instances. Experiments show that our method could get a comparable performance on medium size datasets to existing co-association ensemble method like CSPA or spectral clustering, and is able to handle large scale datasets.