As their backup options multiply, users need to learn from ILM, but at the same time they have to make sure that the ILM and network backup strategies they choose are compatible, as otherwise they could create new problems as they solve existing ones.

So argues Eran Farajan, executive VP at backup company Asigra, whose televaulting software replicates branch-office files over the network and into the datacentre for centralised storage.

"The backup and restore market is diverging, not converging onto a single platform," he says. "The divergence stems from tape being a problem. It solves one problem - backup and archiving - but it creates others. Typically that's slow restores.

"So the market is looking at other options, such disk-to-disk, tape virtualisation and televaulting, or even getting rid of tape altogether and replicating remotely over IP.

"All those technologies are bubbling up to do away with tape outside the datacentre. But there is no convergence - there's no one product to cover all needs, so it will be up to customers to decide what suits."

Key to this decision will be the same information analysis and data profiling that is used within an ILM strategy, Farajan says.

"We like to help customers classify their data into what's critical and what's important, for example by scanning their computers and asking when the file was last accessed, who owns it, what file type is it?"

He warns though that unless backup and ILM are properly integrated - and they might not be if ILM is organised centrally, for example, while backups are done to local tape autoloaders - there can be conflict between the two.

"Customers are investigating ILM whether we like it or not. Whatever ILM solution you buy - EMC, KVS, StorageTek, IBM - they all have one thing in common, which is they use pointers, or stubs. You set up a policy, say to move data after 90 days, and when it moves the data to lower cost storage it leaves a stub behind.

"But during the 90 days when the data was live, your existing software backed it up, so now you have the data on two simultaneous paths, with two sets of backup files."

Not only does that double the volume of backup data you need to store, taking away some of the promised return on investment, but it also creates uncertainty.

"It's one thing to have an ILM solution, it's another to have backup software that's ILM-aware," Farajan says, adding of course that Asigra has an answer: "We recognise the stub file and back that up instead. The software also has to be aware of the several different stub types that operating systems use, such as Microsoft's placeholders."

More generally, he says that ILM could allow organisations to be more creative and economical with their backup policies, whether they use Asigra's televaulting software or existing backup methods.

"Most people would think to put tape into the data centre, and maybe also into the distributed sites. That's the traditional way of doing things, perhaps with autoloaders and appliances in the remote sites, rather than libraries, but still using tape," he says.

"But if ILM takes hold, it only needs to protect your data for the first 90 days, because after that it's on archive tape. So you could perhaps only backup data while it's in the critical state."