Functionality and Performance Evaluation of File Systems for Storage Area Networks (SAN)

The demand for consolidated, widely accessible data stores continues to escalate. With the volume of data being retained mounting as well, a variety of markets are recognizing the advantage of shared data in terms of both cost and performance. Traditionally, common access has been addressed with network-attached fileservers employing data sharing protocols such as the Network File System (NFS). A new approach, poised to deliver high bandwidth access by multiple, heterogeneous platforms to a common storage repository at reduced cost, is beginning to emerge. Storage Area Networking (SAN) is an open-storage architecture designed to eliminate many of the traditional bottlenecks associated with secondary and tertiary storage devices. Conventional high performance computing (HPC) sites and compute-intensive production sites can benefit from such architectures as the need to share computational input and output data sets expands and the mix of computational platforms continues to diversify. Recognizing the potential value of SAN solutions in their overall data management roadmap, the Storage Technologies Knowledge Based Center of the Department of Defense commissioned a research project in mid-1999 to evaluate the functionality and performance of emerging SAN technologies. The initial focus has been on SAN file systems that offer management of disk-resident data. The desire, however, is to expand the effort to include other traditional data storage functions such as backup, hierarchical storage and archiving using tape technologies. The underlying goal is high bandwidth and reliable access to data with guaranteed long-term retention while presenting a seamless and transparent interface to the users regardless of data location. Operational stability and ease of administration are key requirements as is overall data integrity. When complete solutions will be available and just how robust the family of products will be remains unclear. The magnitude of this challenge is realized when considering that production use of these technologies will entail serving numerous, likely