It should come as no surprise to this audience that financial institutions of all stripes — stock brokerages, banks, credit unions and other lenders — are under a certain amount of stress when it comes to data recovery.

These institutions, by nature and by regulation, make multiple copies of daily transactions and the files these transactions affect. Off-site copies are squirreled away for disaster recovery purposes, and near-line copies kept at hand for near-disasters. But the improving performance of data processing systems is making the job of backup tougher, because the capacity to process transactions rises geometrically with every new computer generation and backup systems can’t afford to fall behind.

1996’s methods won’t work

One can’t back up 1998 data with 1996 methods. The arithmetic won’t allow it. Today it is not uncommon for a desktop personal computer to be fitted with a six gigabyte or seven gigabyte hard drive. Workgroup file servers have much larger capacities of course, and data center on-line storage can go into the terabyte (1012 byte) range.

Now consider the Quantum DLT 4000 tape cartridge drive, an example of leading-edge 1996 tape technology. The native capacity for a DLT 4000 is 20GB, and its transfer rate 1.5MB/sec. Simple arithmetic shows that recording a full tape requires 4 hours. Recording two copies requires 8 hours. If one’s transaction volumes are substantially over 20GB, or if one’s cartridge library has been standardized on pre-DLT 4000 tapes, it could conceivably take more than 16 hours to back up 8 hours of transactions — numbers that begin to make conservative managers uneasy.

Faster tape drives are becoming available, but so are larger and faster hard drives for them to back up. And for increased performance, computer system designers have learned to configure disks in arrays (RAID systems) which can record more data more rapidly than can individual disks, making the backup task even more formidable.

Fortunately, certain tape systems technologists have learned to do the same: to configure arrays of tape drives which can record more data more rapidly than individual drives can. They’re called Tape RAID systems. Used tentatively in the mid-’90s, their technology is now proven, and their acceptance growing rapidly.

Mirroring for Multiples,

Striping for Speed

And why not? A two-drive tape RAID system records two copies in exactly the time required for one drive to record one copy. A three-drive RAID, three copies. The technique is called "mirroring," and is ideally suited for fulfilling disaster recovery requirements.

One U.S. west coast financial institution uses mirroring rather more ingeniously for disaster recovery. It feeds backup for two different files to four tape drives, and makes two copies of two files simultaneously.

"Striping" supplies a completely different option. Mimicking what was earlier done with disks, arrays of tapes can be made to operate as one high-speed over-size drive. That is, byte #1 goes to tape drive #1, byte #2 to drive #2, and so on. The result is that multiple drives can be running simultaneously at full speed, and a given amount of data can be recorded in a fraction of the time otherwise required. On top of that of course, two or more times as much data can be stored since there are that many more tapes.

The data is stored across multiple cartridges, although only one copy of the original file is made. However, "striped" recording may also be "mirrored," so that two or more sets of backup cartridges are made simultaneously.

Fail safe

Striping even provides an extra measure of security through the simultaneous creation of a tape with parity codes. What would be four-tape RAID systems are usually 4 + 1, for example. Given an n+1 tape RAID system, any tape drive or cartridge can fail at any time and 100% of the data will be recovered. In addition, the remaining n tapes can be used to regenerate the missing tape, if desired. (The controllers are also smart enough, by the way, to be able to read a set of cartridges which haven’t been put back into the "right" drives, as long as the appropriate number of cartridges have been loaded.)

Other benefits accrue from the fact that a tape RAID system can operate unattended. Today’s tape RAID controllers are compatible with the robotics in tape autoloaders and tape libraries so that operator intervention need not be required for reloading the drives. When operator commands are necessary, as when initiating the backup, a comfortable Windows interface usually is provided.

Oddly enough, tape RAID is a very conservative data recovery methodology. A wide variety of existing tape cartridges and drives may be used. The interface to the host computer is industry standard (SCSI). And no special software is required since the array is made to appear to the host computer like a single drive. An IT organization can use its existing backup software, existing cartridges, and perhaps the drives themselves. Only the tape controller is really different. That, and perhaps the sense of relief.