Linear Tape File System (LTFS) is IBM’s recently released, self describing tape file system that enables tape media to be mounted and read by the operating system when the cartridge is inserted into the tape drive. The goal is to try to make interacting with a tape drive as simple as interacting with a USB stick or external hard disk. The driver is available for Linux and Macintosh with a Windows version coming soon. LTFS requires LTO 5 tape drives and media.
One of the advances made when LTO 5 was released was the ability to partition media; it includes a partition 0 and a partition 1. Partition 0 can be quickly read and is an ideal location to house the directory structure information that LTFS needs to present the contents of the tape media, which is written on partition 1. The result is that a user or administrator can have a view into the media that they can interact with without having to wait for the entire tape cartridge to be scanned. Then a simple copy command can be used to move data between tape and local disk.
It’s important to remember though that tape is still a sequential device and so a copy to or from tape still requires that the media be moved to the correct position by the drive. Depending on the location of the file to be read or where on tape the write will occur, it could take 45 to 60 seconds for the activity to start. With the speed of LTO 5 (140MB/s native) it shouldn’t take much longer to complete.
LTFS is an open source file system software and is available for download at no charge. As stated earlier it’s available today for Linux and Mac OSs with Windows support coming later this year. The first incarnation does not support WORM tape or encryption, features that were introduced with LTO3 and LTO4, respectively, but both are on the road map. Tape library support is also expected from the manufacturers that currently support other LTO generations.
Storage Switzerland’s Take
LTFS is one of the more exciting aspects of LTO 5. Not only does it deliver a very simple way for users to interact with LTO 5 based drives, it also provides a platform for easier interchange of media. Now, at a cost of $50 a TB, 3TBs of information can be overnighted to anywhere in the world. In many cases overnight delivery of an LTO5 cartridge would be faster than transferring 3TBs of data across a WAN segment. The fact that the receiver only needs an LTO 5 drive attached to a computer with the free LTFS driver installed, not a special application, to read that data is a key advantage.
Adoption of this new standard may be different than other attempts at “open tape.” Most of those solutions were proprietary and developed by a vendor that kept the format to themselves, or required the integration of the format into an existing application. By releasing LTFS as open source and enabling its use at a file system level, IBM has removed many of the barriers to its acceptance.
The next big step is to see LTFS used by backup software application companies. While expecting a total conversion to LTFS may be too much to ask, it seems reasonable that there be an option to restore a certain group of files or backups from the proprietary format that the application uses to the LTFS format. Will the application backup software providers make this move? Only time will tell, and the goal might be accomplished sooner by suppliers in the backup virtualization space.
Not only would an LTFS export option be ideal for situations where large volumes of data need to be shared with others, but it could be seen as a best practice for archive data. The advantage that a file system has is that 20 years from now the chances of being able to read it are very good. This has been our argument for disk based archive systems in the past and LTFS tape now would be no different. There is no guarantee that the backup application would be available 20 years from now to read its proprietary tape format. But supporting an open source piece of software, like LTFS, doesn’t require that a backup company remain in business, making the chances much higher that the LTFS drivers will still be available.
Another use is by applications that have appeared in recent years which focus on disk as a backup or archive repository. LTFS now gives these applications the ability to leverage tape as a backup or archive target without having to change their software. The user can now just point at the LTFS mount point and have the application send the data there. LTFS would manage the more difficult task of positioning it on tape. It would be interesting to see these application developers add the ability to categorize and maintain a library of multiple LTFS tapes so they could tell the user which tape is needed and when.
We think that LTFS, if it can gain traction, could be one of the most significant developments in the tape drive space since the introduction of LTO itself. IBM made a very good decision to distribute LTFS immediately and at no cost, instead of waiting for software companies to add their support. If LTFS is successful in this grass roots effort, the software vendors will be quick to fall in line.