System integration and workload migration concerns can complicate the move to an integrated infrastructure platform.
Integrated infrastructures attempt to overcome the integration, management and maintenance challenges that often are encountered in traditional heterogeneous data center environments.
By obtaining servers, storage and networking equipment from a single vendor -- rather than piecing that equipment together in-house from various vendors -- an organization can improve its management insight using tools that are tailored to the specific hardware being deployed. In addition, a single vendor is universally responsible for troubleshooting and maintenance; this eliminates finger-pointing.
However, adopting a converged infrastructure can be problematic in many data centers. This tip highlights the challenges of adding an integrated infrastructure product to an existing data center and outlines potential migration issues when moving from one integrated infrastructure platform to another.
The role of integrated infrastructures
To appreciate the challenges of integrated infrastructures, it's important to define the concept and understand the role of those platforms in an enterprise data center.
The traditional data center is an amalgam of server, storage and network products from a variety of vendors that have been pieced together over the course of years -- usually across numerous technology refresh cycles. For example, it's common to find a data center that uses several different generations of Dell, IBM and HP servers; Brocade switches; and perhaps NetApp and HP storage subsystems. The advantage of such a heterogeneous environment is that IT decision makers can select best-of-breed equipment that is cost-effective and best suited to the computing workloads in use.
But heterogeneous data centers often struggle with systems management that might offer substantial granularity on some systems but not others, forcing administrators to use multiple tools or forego some otherwise important management oversight in the environment. Heterogeneous data centers can also be difficult to troubleshoot. For example, a server vendor may not support a system when attached to certain vendors' networking or storage systems.
Homogeneous data centers overcome some of these problems by limiting the number of hardware vendors deployed in the environment. For example, a homogeneous data center may deploy only HP Gen8 servers or Cisco switches. This can ease management limitations and simplify some troubleshooting, but does not always guarantee proper integration between servers, networks and storage.
The concept of an integrated infrastructure provides data centers with a "one-stop shop" for the entire infrastructure; servers, storage and networking equipment are all bundled and pre-integrated from a single vendor. Current vendors include HP's VirtualSystem, IBM's PureFlex and Cisco's Unified Computing System (UCS). Buying everything from a single vendor eliminates integration and interoperability problems, offers better management because the software tools are tailored for the specific equipment, and service is a single phone call away. It's the ultimate expression of homogeneity.
Before you opt for an integrated infrastructure, however, consider some of the possible downsides.
Conventional data centers and integrated infrastructures
The principal problem with integrated infrastructures is just where they fit with the existing data center. A traditional data center represents a substantial financial investment that organizations are unlikely to discard. Integrated infrastructures are also expensive because they involve a substantial up-front purchase meant to displace much (if not all) of the current data center hardware. This presents a dilemma for companies that are hesitant to replace their existing -- expensive -- infrastructure.
One option is to deploy an integrated infrastructure platform for a second data center; usually a new build (a greenfield project) that has not yet been equipped. In this case, there is no legacy equipment to displace so you're not wasting capital.
Companies that don't have the luxury of having to fill new data center builds with systems may choose to acquire an integrated infrastructure as a prospective technology refresh option -- adding the new platform with the intention of eventually phasing out the traditional heterogeneous infrastructure. When this happens, it's important to consider how to integrate the new infrastructure with the old one.
IT experts such as Chris Steffen, principal technical architect at Kroll Factual Data, said that there is absolutely no substitute for integration testing.
"Have your QA and dev teams beat the hell out of it for a long time first," he noted, emphasizing that thorough integration testing will build experience and quickly reveal pain points -- especially in setup and management. IT teams should avoid infrastructures that simply don't work together.
However, experts such as Pete Sclafani, chief information officer at 6connect, say that current integrated infrastructure platforms incorporate robust application programming interfaces and anticipate some amount of integration with other systems, though proper integration may not be easy or inexpensive. "The challenge is getting development resources to do it when the administrator has other initiatives that may be higher-priority," he said.
"Too many times, systems integration is a 'nice to have' until something goes wrong and priorities or budgets shift," Scalfani added. More frequently, an integrated infrastructure may work with existing management systems for some tasks, leaving more advanced tasks to platform-specific tools.
Migrating one integrated infrastructure to another
Incorporating an integrated infrastructure into your environment causes potential problems, but getting it out of the environment again creates additional issues. This may involve migrating from an integrated infrastructure back to a conventional data center setup or migrating workloads from one integrated infrastructure to another. If you've used an integrated infrastructure platform for three years and the lease is up, how can you move workloads to another platform?
Steffen said to evaluate an integrated infrastructure's off-boarding capabilities early in the evaluation process. There should be a tool or other means to move between platforms -- preferably between different vendors' platforms. If such a tool does not exist, it may be impossible to move to another platform without costly, time-consuming and disruptive vendor-consulting services, which can affect production workloads. This is the worst possible scenario of vendor lock-in.
Sclafani notes that hardware and management challenges are not insurmountable, but a major barrier to switching vendors could lie in training.
"It's tough to invest time, money and personnel learning a new system unless it provides a big upside; especially if the current system is functional," he said, noting it may be possible to upgrade certain parts of the equipment bundle rather than make a wholesale transition. Still, a migration can be more effective and successful once you understand the problem you're trying to solve. "Bundles have many parts in common, even with different vendors, so you have to do your homework on the 'why' behind switching. Otherwise you could end up with the same problems."
Integrated infrastructures promise better management and service for data center operators, but the move to an integrated infrastructure can be an expensive, complicated and highly disruptive one. Prospective adopters must have a clear understanding of how the technology fits into their current data center, how to transition the current data center to the new platform and how to transition away from that platform if circumstances demand it.