Techniques for facilitating and accelerating log data processing are disclosed herein. The front-end clusters generate a large amount of log data in real time and transfer the log data to an aggregating cluster. When the aggregating cluster is not available, the front-clusters write the log data to local filers and send the data when the aggregating cluster recovers. The aggregating cluster is designed to aggregate incoming log data streams from different front-end servers and clusters. The aggregating cluster further sends the aggregated log data stream to centralized NFS filers or a data warehouse cluster. The local filers and the aggregating cluster stage the log data for access by applications, so that the applications do not wait until the data reach the centralized NFS filers or data warehouse cluster.