X
Tech

Spreading your storage eggs

If the UK ever experiences the kind of widespread power-outage seen in Italy recently, mirrored SANs could provide one way to keep data safe
Written by Matt Loney, Contributor
Spreading your storage eggs
Matt Loney
If the UK ever experiences the kind of widespread power-outage seen in Italy recently, mirrored SANs could provide one way to keep data safe

If there is an analogy in IT for putting all your eggs in one basket, the storage area network is it. As storage area networks, or SANs, have grown in popularity over the past few years, enterprises have tended to dump more and more mission critical data into one rack-mounted cabinet.

There are many advantages to SANs, and these include being able to reduce cost and complexity, and easing the management of large amounts of data. Unfortunately, there are also disadvantages. "If you have ten servers in ten locations and one fails, you only lose a tenth of your system," says Dell's European director of enterprise systems, Kevin Libert. "If your servers are all in one location and they fail, you lose everything."

It's a danger that was highlighted over the summer as the power fizzled out across Italy, southern parts of the UK and the eastern seaboard in the US. Such massive, local power failures -- or indeed any disaster -- can mean the lights never come back on for many companies that rely on their computer systems to keep their business going.

Banks are a prime example. At a recent select gathering of CIOs in London, the IT manager of one large city bank, who asked not to be named, cited business continuity as one of his major challenges. In practice, this means mirroring his entire datacentre, he said, not just across town, or a couple of counties away, but in a different country. "I need to be able to mirror my entire SAN across thousands of kilometres," he explained. For banks, and indeed anyone whose business depends on the systems being constantly available, simply backing up is not an option.

Martin Hingley, vice president of the EMEA systems group at analyst firm IDC, says he has seen most of the high-end business continuity work done at financial institutions, but sees more companies taking an interest. "The thing we have seen over past ten years is the concept of failover across mirrored sites," he says. "This started off as something over a campus because of the restrictions, but as the technology has developed, the distances are increasing. In the case of the recent power cuts, if you want to make sure your data remains online then greater distances are essential."

Nortel is one company that has been looking at the network issues associated with mirroring SANs over large distances. Chris Sweetapple, Nortel's director of market development for optical enterprise, says the company started looking at how to take fibre channel outside the datacentre some time ago. "At first we took it across the street to another datacentre, then to another city, in what was essentially a metro-area network," says Sweetapple. But there is more to mirroring a SAN than plugging in a cable, and so Nortel teamed up with Dell and EMC to link two datacentres over 100km, so the companies and their customers could test how well SANs can be mirrored and set to failover across larger distances.

The facility means that customers of Dell, EMC and their partners can test the performance of their SAN mirroring and failover between Dell's Application Solution Centre in Limerick and EMC's Solutions Operations Centre in Cork. Of course nothing comes for free, and the partners expect companies using the facility to end up buying their solutions, which include the Dell/EMC CX600 storage system and Nortel's OPTera Metro 500 equipment.

Dell's Libert says the solution being sold now is synchronous, which means the data is written to the backup SAN first and then written locally before a transaction is completed, "but this means if the other end is much slower then it can drag the whole speed down." In the first quarter of 2004 Dell and EMC plan to launch an asynchronous solution, so that the local data is written first, then cached so the far end catches up when it can.

The distinction between synchronous and asynchronous systems for mirroring and failover of a SAN is important because of its bearing on the performance of the remote SAN that will be required if transactions are not to be lost. For any company investing in just one new SAN -- rather than the more expensive option of buying two new SANs -- next year's asynchronous solution is likely to be the more attractive because it means the current (presumably slower) SAN can be moved to the failover site as a backup. EMC's MirrorView software, which is used to manage the SANs in this setup, is backwards-compatible across previous Dell/EMC product lines, says Libert, adding: "everything is virtual, so if disks are different sizes it doesn't matter."

The interesting thing about the Dell/EMC facility is that it would not, say the companies involved, have been possible in the US. There, the distance between EMC's datacentre in Boston and Dell down in Austin is beyond the capabilities of that country's IP network. But Europe is a different story. "We have largest IP-enabled network in Europe," says Peter Williams, general manager for application hosting at Esat BT, which is telecommunications giant BT's subsidiary in Ireland. "We can sustain gigabit speeds across our network, which is still growing." Right now, says Williams, BT is able to carry up to 12.5Gbps between some 260 cities in Europe. Of course such bandwidth does not come cheap -- costing about 50,000 euros a month -- but few companies are likely to need such high bandwidth, and prices do go down to 4,000 euros a month.

Mirroring a SAN remotely in real time does not come cheap -- it means you need an entire redundant SAN for a start. There are other ways to protect your eggs, but most involve backups and restores. These are simply not an option for a bank, which needs to have its systems online and available 24 hours a day. The same holds true for any e-commerce operation. If it takes three days -- a surprisingly common time-span -- to restore mission-critical data, by the time the business is back up and running it may have no business left to do. At best, the brand is likely to have taken a battering.

Nigel Ghent, regional marketing manager for Northern Europe at EMC, thinks of it is a trade-off between risk and cost. "Ultimately what any business is protecting is its operations," he says. "The heart of it is protecting brand integrity. The EMC logo is probably the single most valuable property we have. With this facility (in Ireland), customers can kick the tyres; can make sure it will satisfy their business requirements."


Editorial standards