Virtualization, automation, and unified computing solutions prompt companies to think about their data center technology refresh plan more frequently than in the past.
In the past, enterprises were leery of revamping their data centers because these projects were complex, difficult to undertake and expensive. But lately the pace of technology has quickened, so they find themselves making adjustments more often.
In 2009, Los Angeles Valley College (LAVC) decided to move from a traditional to a virtualized data center. By using VMware Inc.'s VMware, the academic institution thought it could maximize the use of their data center hardware -- Hewlett-Packard Inc. servers running Linux and Microsoft Inc.'s Windows. For more than a decade, LAVC has relied on HP servers.
"We found that the HP systems were easy to install, and their lifetime warranty helped us guard against any unforeseen problems," said Yefrem Kozin, data center specialist.
During the college's upgrade, HP purchased LeftHand for $360 million in cash. LAVC was already using HP ProCurve switches to support its network, so the next logical step was to deploy HP's BladeSystem Matrix, a converged data center infrastructure solution.
"Previously, our data center equipment operated in separate silos; we wanted to build a data center in a more integrated fashion, because that change would ease our administrative functions," said Kozin.
Vendors, such as Brocade Communication Systems Inc., Cisco Systems Inc., EMC Corp., Hewlett-Packard Co. and IBM, have been building a new generation of data center devices. These products, which go by names such as converged systems, fabric-based computing, converged infrastructure and unified computing systems, consolidate autonomous products -- servers, storage systems, network devices -- into a single system. This emerging computing model promises to reduce costs, speed up deployments and simplify maintenance.
One big potential draw to collapsing server, network and storage systems into a single chassis is cost reduction in areas like physical infrastructure. Companies no longer have to string cable to connect a server or a storage system to a switch, a change that could reduce cabling by as much as 40%. International Data Corp. estimates that as much as 15% of a data center's IT budget goes into the cabling. In addition, fewer devices mean less use of power and cooling systems.
Another plus is that corporations can speed up IT service provisioning. Because the underlying items are integrated, IT staff can get a new system up and running in days rather than weeks, and adding storage or network bandwidth can be done with a few mouse clicks. Businesses are also able to streamline management functions. Since the integrated systems include fewer components, it becomes easier to troubleshoot problem connections.
By the end of 2011, LAVC had deployed its new infrastructure, featuring multiple HP BladeSystem c7000 enclosures, 40 or so HP ProliantBL460/465 x86 server blades, the HP LeftHand P4000 storage system and HP ProCurve 5400zl 10Gbps switches. The new hardware was deployed in two campus data centers and a third remote backup site in downtown Los Angeles.
The college discovered many productivity improvements from the new setup. It streamlined IT management and administrative functions. For example, the time needed to upgrade server disk space went from at least five hours to well under one hour, an 80% reduction. Because of recent technical advances and more cohesion in the infrastructure, the data center now uses less power.
More benefits of a short technology refresh cycle
Capital Region Orthopaedics, a 32-physician clinic, is part of the Albany, N.Y., Medical Center. In the summer of 2010, the company also desired more virtualization and tighter integration among its servers, network devices and storage systems. "We wanted to put in a backup disaster recovery center," said Ray DeCrescente, chief technology officer. In addition, the company was putting in an electronic medical record system, which would have burst the data center's capacity.
Traditionally, CPU cycles limited data center consolidations, but current servers ship with four, six or eight cores per CPU socket, creating plentiful compute cycles. Likewise, server performance can be maximized by layering on virtual machines. However, this new approach leads to new content spots: storage and network traffic often compete for a limited amount of I/O capacity, and the new constraint becomes host I/O.
After evaluating the potential pluses and minuses of an upgrade, the health care provider desired a solution that supported high-bandwidth connections; that lead to the selection of Cisco's Unified Computing System, supplemented with EMC storage systems and VMware Inc.'s virtualization software. The health care provider also examined solutions from Hewlett-Packard Co., which had been its server supplier, and IBM. The health care provider now has deployed a 10 Gbps Cisco Fibre Channel over Ethernet switching fabric.
Capital Region Orthopaedics, which spent $2 million on its new system, now has about 75 virtual machines supporting 52 TB of information. By making the change, the health care provider found its administrative tasks reduced. New virtual servers are allotted with the proper amount of bandwidth and storage through a central console. The new system also includes improved diagnostics that enable the health care provider to ensure system availability. "We could not move to an all-digital environment without first deploying a more flexible computing infrastructure," said DeCrescente.
Recent technical advances, like virtualization, are prompting enterprises to take a close look at their data center infrastructure. While movement to unified computing solutions can cost millions of dollars, many enterprises think such investments are worthwhile.
About the author:
Paul Korzeniowski is a freelance writer who specializes in data center issues. He is based in Sudbury, Mass., and can be reached at firstname.lastname@example.org.