Managing the data we store in IT systems is becoming as complex as the data itself. Adding to the problem is the increasing tendency for most organisations to hold on to all the data that comes into their possession. There is also a trend for legislatures worldwide to mandate organisations to retain more data - down to email messages and digital recordings of phone calls to and from the organisation. John Costello cuts through the complexity and illustrates what is happening in data storage systems.
The blurring of the lines between companies recognised as specialists in data storage products and mainstream IT companies is adding to the complexity for most IT managers. In the first group we could quickly add companies like EMC, Storage Technology and Hitachi. In the second group are the familiar names of IBM, Hewlett-Packard and Sun Microsystems. Just how blurred the lines are becoming became obvious when researching this article. EMC chose to highlight its VM Ware product which has little to do with storage but more to do with overall management of system resources.
It's a continuing trend, according to EMC system integrator, Enstor.
Ken Wood, Enstor's marketing director, says more of EMC's revenue is coming from software products such as VM Ware. "Data storage is much more than spinning disk platters. You also need software to manage the data." Kevin McIsaac, from research company Meta Group, says companies can now look to spend as much or more on software to manage data as on the physical hardware storing the data.
"Hardware costs are declining by about 35 per cent each year," McIsaac says. "There are also hidden traps in some software products - such as being charged for installed capacity rather than actual usage," he says. Overall, there is a trend for more organisations to centralise data storage to reduce costs and maintain control over data.
In a survey of medium and large Australian organisations, Storage Technology (ST) found more than two thirds were either planning (32.3 per cent of respondents) or considering (38.5 per cent) consolidating or relocating more server and storage infrastructure into a data centre. The survey covered 138 organisations across Australia.
When asked about the major considerations driving the trend to data centres, reduced costs followed by better operational control were cited as the major reasons. More than half the respondents said they were also looking for improved protection for data with the move to a data centre.
Typical of such organisations is Barwon Water. The utility provides water and sewerage services to more than 250,000 people in Victoria centred on the city of Geelong. While reduced costs was a factor behind the decision by Barwon Water to consolidate data storage, the organisation also saw it as making it easier to deploy new applications.
"We had a situation with islands of data storage and a mixed operating environment - Solaris and Windows," says Joe Adanski, executive manager strategy and technology for Barwon Water. This will be consolidated into two centres with Hitachi storage systems. The centres will be linked by fibre optic cable. Also reducing costs is the Commonwealth Bank. In a major storage consolidation program, the bank is reported as saving millions of dollars in equipment, administration and maintenance costs over just six months.
It replaced 30 file servers and accompanying tape drives with two disk-based systems from Network Appliance. Network Appliance has also recently signed up Telstra and Department of Defence as customers.
Managing data stored in a single facility means separating it into tiers according to its importance. Many vendors give this the fancy title of information lifecycle management.
José Goldmann is principal of SLI-Consulting, an independent consultancy specialising in data storage. "Gartner Group reports that more than 80 per cent of corporate data is never accessed again after being stored for more than seven days," he says. "Placing this data on a corporate storage platform is not only inefficient it is very costly. Organisations need to match their storage needs with their business needs for maximum efficiency in data storage."
Goldmann says the answer is to segment the data into tiers according to its importance. In the top tier is mission-critical data such as financial and production information. In the second tier is business-critical data such as messaging and office applications. Finally, there is what he calls standard data - reference information and point-in-time copies.
Fundamental to the segregation of data is an organisation's ability to understand the nature of the data being stored says Ian Selway, marketing manager for Hewlett-Packard's StorageWorks product line. "They also need to be able to understand the changes taking place in their environment and the effect this has on the data they are managing," he says.
While the ST survey found a move to data centres, it also found about 40 per cent of all organisations said they had some mission-critical data not stored in a secure data centre.
But according to David Cowell, ST's Asia-Pacific manager for data centre services, this only tells half the story.
"The data centre needs to be better in terms of risk management than the locations where the data is being moved," he said. He says this means an organisation has to weigh up the risks and costs of down time. "If you are aiming for 99.999 per cent up time for a year it means you are prepared for total annual down time of between one and two minutes," Cowell says. "If you are prepared to accept 99.99 per cent then your annual down time is about 20 minutes and for 99.9 per cent means you are looking at three hours annual down time.
He also says about half of all organisations do not have a data centre. "Data is stored on office servers or even individual PCs," he added. "This is also likely to see multiple copies of the same data stored in different locations. "In effect you are managing redundant data."
Once your data is safely consolidated in its new data centre, there may be other problems.
According to Leanne Cunnold, Australian general manager for American Power Conversion, heat related failures are becoming the major cause of network downtime.
She says cooling systems in data centres are room- or rack-based. Typical cooling rates capable of handling about 3 kW per rack may not be enough for racks using blade technology that require as much as 30 kW.
Likely to accelerate the trend to consolidating data in data centres is an increasing amount of legislation requiring organisations to store even more of the data they acquire.
In the past two years, Cowell said he had not come across one organisation that had deleted any data in its possession. This has led to another of the IT industry's catch phrases - information life cycle management.
|
More bytes than you can put a name to One of the minor issues about data storage is dreaming up new terms for ever larger numbers of bytes under our control and flowing over our networks. Measuring data storage in megabytes is almost a distant memory. The next measure was gigabytes but that is now giving way to terabytes. The tera-prefix was as far as the metric system stretched when it was first introduced to Australia in the early 1970s. SI, the international body responsible for setting standards, introduced the peta prefix in the mid-1970s. A contraction of the Greek word 'penta' meaning 'five', a petabyte of data (2 to the power of 50 or 1,125,899,906,842,624 bytes) should be enough for anyone to manage. For some organisations the terabyte is sufficient. For example, the Australian Tax Office has about 70 TB online. But terabyte or even petabyte isn't big enough for all of us. So we now have 1000 petabytes equalling one exabyte. Waiting in the wings are vottabyte (1000 exabytes) and zettabyte, (1000 vottabytes). Putting this into perspective, one of the major TV networks in China is planning to convert its entire library of analog material to a digital format. Involved is about 800,000 hours of video. It will need an estimated 10 exabytes of storage. And yes, they have thought about the next prefix after zeta. One suggestion was to call it a stevi - after Simon Stevin. He was the Belgian mathematician who first proposed the idea of a metric system. That was back in 1585 - roughly 13 teraseconds ago. |
Hewlett-Packard breaks the life cycle down into five distinct stages to help organisation better manage data. Andrew Manners, HP's director of network storage solutions labels the stages: create/modify; replicate and distribute; archive and recall; protect and recover; and remove. "We address each of these stages by managing information based on service level objectives defined within the business, as well as ensuring that business processes are aligned," he says. The increasing complexity of managing data has seen the rise of companies finding specialised niches in the market.
Brocade Communications Systems labels itself as the world's leading provider of storage area network (SAN) infrastructure solutions. It supplies its data storage switching technology to most major storage system manufacturers including IBM, Hewlett-Packard, Sun Microsystems and Hitachi. It is scaling its technology down to reach small to medium business organisations (SMBs).
"We are working with our partners such as HP and IBM to bring SAN technology down to the SMB market," said Graham Schultz, Brocade's manager for ANZ. Brocade recently launched eight-way and 16-port switched, which it says will reduce the cost and complexity of deploying SANs in SMBs.
The company sells exclusively through OEMs, resellers and system integrators.
But Schultz recognises that an enterprise-wide SAN may not be suitable for all organisations. "We can see multiple islands of SANs in some organisations," Schultz says. "There are probably many reasons - including political ones - where you don't want these SANs connected," he says. "But they have to be managed." Brocade claims to have more than 63 per cent of the total SAN infrastructure market.
Sun Microsystems sees one solution in resolving the complexity issue is the move to a utility model. "We see the idea of buying storage units one by one as storage needs increase is not the way to go," says Dan Kieran, storage business manager for Sun.
"We're looking at working with our customers on a three-year plan to manage storage," he says. According to IDC, Sun is third on the list of companies in terms of total revenue from storage products. Ahead of Sun is leader HP, followed by IBM. Just behind Sun is EMC.
One question that regularly cropped up was just how much more data can we cram onto spinning disks. Ron Nieboer, a storage strategist with ST, says disks have a life of between three and five years. "Newer systems automatically copy from disk to tape and tape has a proven shelf life of decades," he says.
The capacity of the standard half-inch (1.27cm) tape cartridge has been steadily rising. When launched in 1987 it had a capacity of 200 MB. It is now 200 GB or up to double that if the data is compressed. "That's a thousand-fold increase in 17 years but I expect that rate of change to be even more dramatic," Nieboer says. "Next year I expect to see capacity of the standard cartridge increase to 1 terabyte (TB)," he says. "But we are creating new kinds of data - look at how much space multimedia requires.
You don't need to be a genius at maths to work out how much space you'll need to digitally store all the two-way traffic in (say) a 100-seat call centre, for example.
Assume your 100-seat call centre operates at the equivalent of full capacity for 10 hours a day. We'll assume operators and callers talk at the average rate of 90 words a minute.
The average word needs around 64 bytes. Do the maths. In one day you'll need to record 346 megabytes of data.
You just can't keep storing that stuff on disk," Nieboer says.
