Trillions of bytes were transferred to tape last Friday. Trillions more will be transferred next Friday as thousands of full weekly backups are run. Everyday autoloaders click into action and billions of bytes pass into them to be stored. At the end of the week trays of drives are taken away and stored in case they are needed.

In tape libraries large, small and massive, millions of tape cartridges are stored with robotics ready to pick the ones needed out and transfer them to drives for files to be restored or added to.

In smaller businesses individual drives are used for smaller backups.

All around the globe there are billions of tape cartridges sitting in silent vaults ready to be used if disaster strikes. It is probably the world's single biggest data loss insurance activity. There are probably more people involved in tape backups and possibly more money invested in it than in any other data loss activity, possibly in terms of any insurance against loss activity by the owners of the data.

The disasters such a gigantic world-wide effort are focused on are of just two types, the one including the other.

First there is the danger of a disk crash. This will happen. Your disks will fail sooner or later. You cannot rely on disks not to fail. When one crashes then, for all practical purposes, the data on it is gone, lost for ever. You have to get it back if you can, and you can if you have stored in a tape archive.

Secondly there is the danger of disaster, of your whole computer system being rendered useless by a disaster. These things happen. Ask the business residents of New Orleans. Nine eleven no longer means only a fast Porsche car, not since the World Trade Centre twin disasters.

Generally speaking, with proper IT discipline, tape backup works. There are firm processes involved, processes that must be run. You can't simply assume backups complete when left alone, and you can't assume a tape cartridge lasts for every. It needs checking.

But tape is invaluable, irreplaceable really. Nothing else could hold the vast volumes of data required for so little money.

However, it is being asked to do a job it wasn't meant for. It's being asked to recover files that careless users deleted. A backup archive comprising millions of files relating to an entire business's operations, representing millions of pounds of invested capital and revenue is being asked to retrieve an inadvertently deleted mail message relating to the purchase of a pair of door hinges, or something equally trivial.

Of course it isn't trivial to the distraught owner of the missing e-mail. And there's no other practical route to getting the lost file back. We know he only wants the equivalent of a tiny AA battery's worth of power but, even so, we have to switch a nuclear power station on to satisfy the request. We know it's ridiculous.

The disk alternative
Now, with cheap serial ATA (SATA) drives and RAID, you can back data up to disk. You could keep a week or a month's worth on disk and then back it up again to tape, the only practical bulk data store; disk is still just too expensive. But disk is great for responding to the low level, end-user lost-file restore requests. It's much faster than tape; disks ae permanently mounted, and their random-access nature means that the time to locate the file is fractions of a second versus seconds, possibly minutes on a long tape. Disk-based restoring of data can be ten times faster than tape.

Even better is that the disk backup can be represented to users as the familiar Windows explorer interface. They can find their own files or mail messages and then drag-and-drop them to their own system's folders. It doesn't need to involve the IT department at all, beyond having the disk backup there is the first place.

Most restore requests come within a week or so of the data being backed up. So disk is better than tape in user responsiveness terms for this kind of data backup and restore. Another benefit is that backup time shrinks hugely. You can get data onto disk much faster than tape. IT departments suffering from backup window congestion can get relief by interposing disk-based backup between their servers and the tape systems.

But that doesn't mean disk is going to wipe out tape. Just because you can use your existing tape backup software to backup to a virtual tape library (VTL) on disk - ask HP about its Sepaton ('No tapes' read backwards) VTL - faster than to a real tape library, and have end-users doing their own restores doesn't mean that 20TB of disk backup isn't a whole lot more expensive than 2-TB of tape backup.

There is a cross-over point between the costs of the two media. But there is another consideration apart from cost and speed and end-user convenience.

Tapes can be dismounted and taken off site. Disks used for backup cannot. If your data centre gets flooded down go your disks, backups and all. The only way you can have disk-based backup and disaster recovery is to have your backup disks off-site. That means backing up over a network link; extra cost, or physically removing disk drives.

Disk drives are fragile things and you may not want to do that.

Who to talk to?
The disk-vs-tape backup issue has quite a few serious factors to consider. Who could you talk to about the relative merits and places to use disk and tape backup? Storage Expo is full of good companies to discuss it with and explore supplier's attitudes. Go and have discussions with COPAN Systems and ask it what MAID does for backing up data. Talk to Bridgehead Software, Dot Hill, EMC, H-P, HPS for a system house view, IBM, Network Appliance, Overland Storage, Quantum

There are other alternatives to tape backup as well as disk - online vaulting for example. Talk to EVault about this at the show. There is also archiving to optical disk. Try Princeton Softech and CommVault for that, also, of course, Plasmon with its UDO format, and its LTO Libraries.