Time was when the end of the working day meant that IT systems could stop serving users and begin precautionary activities such as back-up. Globalisation and the march towards online business has meant operations must be continuous, leaving no downtime in which to back-up.

This absence of a back-up ‘window’ has changed the way businesses think about protecting their data from accidental loss or corruption. In some cases, business leaders can be faced with the prospect of having to cherry-pick which types of data get backed up, says Chris Sweetapple, general manager for Europe, the Middle East and Africa (EMEA) for back-up vendor, Avamar. “If there isn’t time to do it all, it can seem like there is no other choice.”

However, others have turned the thinking about back-up on its head. Instead of viewing it as a lengthy operation conducted once a day, in which all data is committed to tape, back-up is increasingly seen as a continuous process in which incremental changes are committed to record as they occur. 

But this drive towards real-time data protection has other drivers too. “Businesses are saying, ‘When disaster strikes, we want to lose less data,’” explains Dave Russell, research vice president at analyst group Gartner.

That desire is usually expressed as a ‘recovery point objective’ (RPO) – the time, relative to the disaster, at which it is possible to recover data. For today’s business leaders, a RPO of a day could represent the loss of thousands of financial transactions or the breakdown of a critical process.

A third impetus for adopting more sophisticated back-up systems is the need to restore systems faster should the disaster recovery plans need to be invoked. In the world of back-up jargon, this is referred to as the recovery time objective (RTO).

Continuous back-up

An emerging approach to these difficulties is continuous date protection (CDP). CDP software replicates every single change in a database as it is made, without interrupting queries to that database. The data can then be ‘rewound’ to any number of restore points depending on the available storage capacity. 

To date, CDP has been seen as a solution reserved for file and print data because applying replication to multiple applications, in which the same data occurs at many different points, raises questions about which instance of the data should be regarded as the master.

That is changing however, with companies such as XOSoft and Mendocino offering CDP software that can be applied to enterprise applications.  Some indication of the viability of these tools can be seen in the acquisition activity surrounding the technology; CA acquired XOSoft in July 2006, and storage giant EMC picked up CDP provider Kaysha in May.

Nigel Tozer, principal consultant at CA describes CDP as a way of protecting individual users’ data as they use applications in their day-to-day work. “The ability for end users to rewind their application data to previous point has an enormous impact when scaled up to the entire enterprise.”

And Gartner’s Russell believes that the appetite for CDP will increase as users’ expectations for information availability advance. “I think in time everyone will expect zero loss of data.”

But he also adds that no enterprise yet relies on CDP as its principal back-up provision. That is because of the large amount of disk storage it consumes to preserve a few hours’ activity. In the data centre, he says, businesses must weigh up the frequency of available restore points with the cost of storing multiple replications of so much data.

Time pressures

Managers at UK publishing house Johnston Press have been grappling with the conundrum of backing up data in a fast-growing organisation.

Business has been booming at Johnson Press, thanks in part to its numerous websites, set up for readers of its local newspapers. The business is built on drawing in readers through compelling content, and then advertising to those readers. With over 300 portals for its local newspapers, it is producing vast swathes of data.

This exponential growth in the demands on the content management system has meant that the time reserved for back-up, between midnight and eight in the morning, was being squeezed down to an impossibly short period. 

And as the online business grew in financial significance (revenues have grown by a factor of 20 since 2000), it has become crucial that news stories posted on websites can be restored as quickly as possible should the data be lost: after all, there is no point reverting to yesterday’s out-of-date news.

This drove the investment in an up-to-the-minute back-up facility, built in conjunction with computer maker Dell. The company established a business continuity volume that replicates any changes made to the data. On top of that, snapshots are made of that volume every two hours.

This means, firstly, that the business can always restore to the point just before an incident of data loss or corruption, and secondly, that back-up does not drain the resources of the operational systems, as it is the business continuity volume that is backed up. “We go back to the old data at least once a month,” says Dave Martin, group technical and development manager of digital publishing at Johnson Press.

And it is not just data volumes that add to back-up difficulties. For highly distributed businesses, where as well as a main data centre, branch offices will produce business-critical information.
 

Cherry picking

One of the challenges posed by the branch office structure is related to the distance from the back-up site. If businesses are providing a continuous stream of data from branch office to the back-up site, there can be an issue of data fidelity – ensuring that the information is recorded in the correct ‘order’, so that an accurate record of changes can be made.

“Over long distances, this causes a problem,” says Boaz Palgi, director for EMEA, of back-up software maker Topio. The traditional approach has been to do synchronous replication, but this is expensive, although new technology promises to be able to do the replication asynchronously and cheaper, he adds.

When it comes to the expense of back-up and disaster recovery, business leaders have some hard decisions to make, says Palgi. “Ideally, everyone want an RPO of zero, an RTO of zero and the price to be zero.”

In reality, companies need to work out where they cannot afford any downtime, where they cannot afford data loss, and whether they can afford to allocate the necessary bandwidth, says Palgi. That requires some form of data classification.

In effect, the windowless back-up phenomena is forcing companies to rationalise which parts of the business information receives priority. Some see this as cherry picking, others prefer to regard it as back-up optimisation.

Further reading

• More articles from Information Age's Storage Business Briefing.
• For more stories like this, check out the Business Intelligence briefing room.