Data storage: It’s a strategy not a technology
By Paul Demery
It seems like a dry area—data storage. But it’s crucial to retailers’ operations,
and lacking a well thought-through strategy it’s an area where costs can spiral
out of control without anyone even noticing what’s happening. “Utilization of
storage systems is often 50% or less,” says Phil Goodwin, senior program director
and specialist in storage technology for research and analysis firm Meta Group.
“In a lot of cases, storage efficiency is so low, clients don’t have tools to
determine what their storage growth requirements are. So to avoid problems,
they buy far more than they need.”
Because not all stored data are the same, data storage is not just a technology,
it’s a strategy, particularly in a data- and time-sensitive industry like retail.
But determining which data to store where in the most cost-efficient manner
continues to elude companies. Some retailers stick with a single-tier storage
system for all types of data, an option that often results in wasteful use of
storage infrastructure. Even worse, such an approach makes critical data that
retailers could use to make split-second decisions inaccessible. “Because today’s
retail environment is highly competitive, there are greater response-time issues
for access to information and the time to react to demand has been reduced,”
says Gene Alvarez, retail industry analyst and vice president of technology
research services at Meta Group.
Rocketing growth
But even as many companies have not determined the best ways to use data storage
systems, demand for data storage is surging. “Demand is growing at a 50-55%
compound annual rate,” Goodwin says, adding: “Average storage expenses today
consume 12-15% of information technology budgets, but we expect that to increase
to 15-18% by 2006 or ‘07.”
Aberdeen Group estimates that the data storage market, covering total investments
in data storage systems, will grow more than 40% in three years to more than
$100 billion in 2005 from less than $70 billion in 2002.
Some retailers with sufficient awareness of data storage issues maintain separate
sets of storage systems for different data categories—maintaining a faster storage-and-retrieval
system for POS data, for instance, than for information on customer loyalty
programs. As competition in the retail industry forces merchants to be more
responsive with data—for example, to check a customer’s purchasing activity
to recommend additional sales—they’ve come to rely more on multi-tiered storage
systems that cache the most critical customer-serving data close at hand for
speedy storage and retrieval. “It’s become more crucial to cache your most critical
data on your higher-end storage servers,” Goodwin says.
The results of an effective data storage strategy can significantly reduce
operating costs while improving a retailer’s ability to compete with information
that enhances the customer’s buying experience, experts say.
At German grocer Edeka, direct queries to its data warehouse are 50% faster
under a new configuration that includes multiple levels of storage from IBM
Corp. “That allows us to forecast demand growth and suggest corrective measures
in the event of a sudden surge in returned goods, for example,” says Sven Hohmann,
Edeka’s manager of data warehouse solutions. Edeka uses IBM’s iSeries data back-up
system with IBM’s TotalStorage Enterprise Storage Server for high-end POS and
inventory data. “Retailers need to know when something has happened to demand,
either up or down the supply chain, and they need to pass that information to
their supply chains,” Alvarez says.
And the first retailer to access and process that kind of information is most
likely to win out with increased sales and improved customer service. “If business
is ratcheting up, for instance, merchants need to quickly buy more goods,” Alvarez
says.
The Internet plays an important role in making data storage systems more effective
by providing broad-based access to stored data via web browsers, experts say.
“I see more and more web-based systems,” Goodwin says. “It’s becoming a de facto
standard to have a web-based storage system, because if a storage system is
geographically dispersed, it can be controlled from a central point with web
browsers. The web improves accessibility for users. Even though storage may
be dispersed over an enterprise, it can be controlled from a central point.”
One of the biggest changes in data storage has been the realization that the
strategy that works for one retailer will not necessarily work for all others—even
any others. “In the past, most major retailers took a one-size-fits-all strategy
to storage,” says Ken Steinhardt, director of technology analysis for data storage
systems provider EMC Corp. And to assure that all data had sufficient storage
capacity and server performance and reliability, companies would often over-invest
in high-level storage technology, he adds. “Almost everything was stored on
the highest class of storage, offering the highest performance, availability
and functionality—but also the highest price,” he says.
Matching tech and data
Retailers today need to better match their data-storing needs with technology,
adds Goodwin. “They don’t have a systematic methodology for matching service
level requirements with storage infrastructure,” he says. “So in many cases
there is room for improvement. A lot of organizations on a one-tier system could
be spending more on storage infrastructure than they really need to.”
For example, he says, data that don’t need to be accessed quickly can be stored
in lower-priced systems with the same level of data integrity as more expensive
versions. “You can get similar reliability in lower-end systems, not the same
performance in terms of speed, but the same reliability in terms of redundancy
of data,” Goodwin says.
Furthermore, better understanding of data storage needs can result in such
compelling cost savings that a retailer would be well served to invest the effort
in breaking down its storage needs and deciding the time-sensitivity and importance
of all its data elements. “I don’t see any reason for retailers to hold off
at this point,” Goodwin adds. “Going to a 2-tier data storage model can save
75% on hardware and software costs, and going to a 3-tier model can save 21%.”
The three categories
Goodwin places stored data into three categories, based on descending order
of importance. The highest category, which he terms mission critical data, is
information that must be recovered within a short time frame—say, four hours
or less, so as not to disrupt operations critical to a company’s success. The
two most important examples for retailers: POS data and inventory levels, information
crucial to process orders and serve customers without losing sales.
The second category is business critical information, data crucial to operating
a company on a daily basis but that the temporary loss of which won’t result
in lost revenue or sales. “Back-office financial accounting data is business
crucial; you’ve got to have it, but if you close the books today or tomorrow
it’s not as important as keeping your POS operation moving,” Goodwin says. Business
crucial data, he adds, might be set in storage for a maximum retrieval time
of 12 hours.
Less important data—-which could entail customer loyalty program information,
product lifecycle records or old POS data kept for long-term backup—constitute
the third tier and might be set for a retrieval time of a few days.
Within the data storage vendor community, Goodwin notes that the high-end
servers with the fastest data transmission performance and deepest reliability
are Hitachi Lightning, EMC Symmetrix, Hewlett-Packard EVA and IBM Shark systems.
He places within the mid-range of storage servers EMC Clariion and IBM FastT,
and within the low-end, Dell Power Edge, EMC Sentera — though he notes that
Dell is also moving into the mid-range.
In addition to segmenting of data by importance, vendors are developing other
ways to save money with data storage systems. Acxiom Corp., for instance, offers
a system of computing power from a grid of computer processors it maintains
for clients. In the second quarter of this year, it expects to make available
a grid-based storage system that clients can tap. By using low-cost components
in a grid environment that mirrors data for reliability, Acxiom says it can
provide data storage at about $5,000 per terabyte of data, compared to $30,000
to $40,000 per terabyte for more conventional storage systems. “This will make
a lot of storage available that might otherwise be cost-prohibitive for some
retailers,” says Alex Dietz, products organization leader.
While investments in data storage systems often favor a single vendor platform
to assure the best integration, the evolution toward open architecture and common
industry standards is making it easier to for different vendors’ systems to
work together. “Most of the differences in the market have disappeared,” Steinhardt
says.
Easier connections
That stronger integration is also making it easier to connect storage systems
to sources of data. “It’s very much plug-and-play,” Steinhardt says. “We make
products now so you don’t have to make changes to an existing application server
environment.”
But retailers still need to conduct due diligence in checking different network
components to assure a smooth flow of data. “You have to look at the entire
spectrum of the IT solution and match processors with storage systems and communication
networks, because the system will never be faster than the slowest bottleneck,”
Steinhardt says.
paul@verticalwebmedia.com
Click
Here for Data Management Solutions
