Media Coverages
KPMG ANALYSIS
Networks, Data Centers Evolving Toward Flexibility
By Dave Pelland, Managing Editor, Technology Insider
Server virtualization and the automation of management software tools are helping data centers evolve toward fully automated facilities that are easier to configure and operate.
"You're going to see a continued push on lowering the cost of computing," said Mark Hurd, chief executive and president of Hewlett-Packard Co., at the Gartner Symposium/ITxpo in Orlando. "Ten years from now, data centers are going to be [unmanned], driven by software automation tools and management software."
Hurd said specialized applications such as storage management or systems management software are likely to come together into unified programs.
"Instead of all these point products or even point companies, [management software will have] broader enterprise capabilities that can be linked as these technologies come together," Hurd said.
As companies look for more flexibility from their data centers and networks, several trends are expected to converge over the next five years. Instead of buying servers dedicated to specific functions, companies will purchase commodity hardware that's easy to reconfigure.
Thomas Bittman, a Gartner vice president, said that companies want to link their information systems with those of major customers and suppliers to reduce operating costs. Companies also want to increase the agility of those systems, so they can respond to changing market conditions without replacing major applications or hardware.
"[A] business needs to react quickly, and [its] infrastructure needs to adapt," Bittman said. "IT is shifting from this fixed-cost thing you change by buying more equipment into something that needs to be more flexible."
Gartner vice president Donna Scott said more companies are consolidating servers and turning to "virtualization" -- using software to separate computing resources such as processors and memory from physical devices -- to reduce the cost of server maintenance.
"A smarter way to build systems is to allocate workloads dynamically and adjust them when business needs change," Scott said. "Companies want the ability to use automation to bring components into production environments as they need to."
Virtualization is a shift from buying equipment from specialized vendors and allocating it to specific tasks. Applications such as e-mail or displaying Web pages can be combined on a single physical computer that is partitioned to act as separate devices.
"You can have one server act like a bunch of different boxes, or you can aggregate the capabilities of different boxes so they look like a single, larger server," said David Cearly, a research vice president at Gartner. "Being able to aggregate up or slice [servers] down gives you great flexibility, and you can ignore the physical limits of what you happen to own."
Cearly said that if a single physical server is operating as three or four virtual devices, the workload can be shifted easily from one device to another. If a virtual server has a software crash, for instance, its work can automatically be moved to a device with enough capacity to handle it.
"You can move an application on the fly to another box, and the users won't notice," Cearly said. "If you need to update a server, you don't have to wait until the weekend [to avoid disrupting users.]"
Along with its flexibility, proponents of virtualization say the practice helps customers avoid being locked into equipment or software from specific vendors as their needs change.
Over the next five years, virtualization and management software are expected to converge into policy-based automation, according to Gartner vice president Steve Prentice. Over time, companies will be able to connect a server to its network and allow software to find the server, determine its capabilities and assign it a needed computing task.
If that server eventually crashes, the network will detect the loss and attempt to reassign its work.
"Infrastructure will increasingly be delivered as a series of services," Prentice said. "This will result in [companies having] a very thin IT organization that's responsible for strategy and coordination."
For instance, companies may feel that their order-processing database is more important than other applications. If order-processing performance starts to degrade, the other applications can be slowed, and the network can transfer additional computing capacity to improve order-processing performance.
"Companies want the flexibility to sometimes have servers operating at 2 percent of capacity, and other times running at 99 percent," said Robert Gardos, CEO of GridApp Systems. "They want to manage it centrally, so if their capacity requirements change, the system should be smart enough to adjust and make the needed changes in real-time."
But Gartner's Scott cautioned that full-fledged data center automation remains a goal to be addressed over the next five years.
"A lot of this is not real yet, and there's a lot of vendor hype out there," Scott said. [Customers] have to be wary because this kind of infrastructure is not yet easy to achieve."