GridFS: Ensuring high-speed data transfer using massively parallel I/O

I/O has always been performance bottleneck for applications running on clusters. Most traditional storage architectures fail to meet the requirement of concurrent access to the same file that is posed by most high-performance computing applications. While many parallel and cluster file systems meet this requirement, they are still plagued by metadata overheads and associated management complexities that prevail in read/write intensive scenarios. In this paper we introduce GridFS, a next generation I/O solution that can scale to hundreds or thousands of nodes and several hundreds of terabytes of storage with very high I/O and metadata throughput. It is besed on Object based Storage Architecture (OSA) model and goes a step further to eliminate runtime file access overheads as compared to other implementations on the same model. By eliminating most access overheads and optimizing metadata, GridFS outperforms other solutions in read/write intensive scenarios and this makes it better suited for I/O intensive applications like seismic analysis, weather forecasting, genomics and 3D/4D design simulations.