Protecting File Systems : A Survey of Backup Techniques

This paper presents a survey of backup techniques for protecting file systems. These include such choices as device-based or file-based backup schemes, full vs. incremental backups, and optional data compression. Next, we discuss techniques for on-line backup (backups performed while users continue to access the file system); these techniques include file system locking and creating instantaneous, copy-on-write “snapshots” of the file system. Last, we discuss protecting data from site disasters and media deterioration. The paper then classifies several research and commercial backup systems according to the parameters already described. The classified systems include the UNIX dumpand tar utilities; hierarchical storage managers such as IBM's ADSTAR Distributed Storage Manager, Legato's NetWorker, and UniTree; and several research and academic systems. We conclude with measurements of full and incremental backups over a one-year period in a large networked UNIX environment at Georgia Tech's College of Computing.