Victoria - Fotolia

IT shops share hyper-converged infrastructure experiences

Converged and hyper-converged infrastructure users say the bundled architecture changes daily operations in the data center.

Converged and hyper-converged infrastructure systems are increasingly deployed for enterprise workloads, changing how IT shops perform data center IT management and expansion tasks.

The range of IT shops moving to converged infrastructure (CI) systems varies from those at small and medium-sized businesses to midmarket enterprises with limited IT resources, in all verticals, said Christian Perry, senior analyst at Technology Business Research (TBR) in Hampton, N.H. Because converged infrastructure -- a bundled architecture of compute, storage, networking and management software -- eliminates a great deal of IT work on integration and maintenance, it is often favored by the business managers who require more resources.

The difference post-convergence

IT shops that have spent months and even years on hyper-converged and converged infrastructures experienced significant changes from legacy infrastructure.

"Due to business changes, the IT team had gotten smaller, and we were understaffed," said Steve Schaaf, CIO of Francis Drilling, a transportation logistics services provider based in Lafayette, La. He turned to SimpliVity OmniCubes in part because they required no investment in training or IT reorganization.

Schaaf tracked converged technology from its inception with Cisco UCS, but waited for a refresh cycle to purchase OmniCubes to replace IBM blade servers and a storage-area network (SAN) array that was at capacity.

"I don't have to calculate how much my SAN can handle [anymore]," said Wesley Kennedy, systems engineer at Carenection, which provides private ISP for hospitals. Carenection started with one Nutanix hyper-converged box in its Columbus, Ohio data center almost three years ago, adding a second about 18 months later in its active backup facility in Las Vegas.

To add a networking port with traditional infrastructure, for example, Kennedy's IT team verified that enough bandwidth was available via top-of-rack switches, that the iSCSI and production and DMZ networks remained isolated, and that they followed extensive change control processes. With Nutanix boxes, the virtual local area network just has to connect into the right port.

"When I get two more nodes, by the time I get to the data center, the ports will be ready for me," he said.

CI also reduces maintenance and troubleshooting efforts to cut costs from the operational IT budget.

"Our data center is dark most of the time [because it's] reliable," said VCE vBlock user Michael McGlade, systems and operations manager at A. Duie Pyle, a midsize, regional trucking company that covers the northeastern U.S.

Shedding a heterogeneous IT equipment mix to centralize on one vendor's product -- often considered a drawback of CI by hitching the IT cart to one prize horse -- pays off in higher visibility into the infrastructure.

"We can implement a chargeback program for business groups on [the OmniCubes]," Schaaf said, because of the detailed management system.

The level of integration throughout the application's platform comes at the expense of bleeding-edge operations, McGlade said.

"You're not going to get the latest greatest software releases the day they come out," he said. "The vendor is going to do the testing and integration with the platform first."

CI expands the IT empire

IT teams expand capacity for two reasons: gradual demand increase and new projects. Both are true of converged infrastructure users.

"VDI was always on our roadmap, but we held back because of the storage requirements, as well as the server resources needed for it," Schaaf said. Now, Francis Drilling will plug in another OmniCube to run VDI. Similarly, the company hosts disaster recovery on OmniCubes in its backup facility.

The converged system was a better option for expansion than cloud because Carenection doesn't need bursty capacity, Kennedy said. The company added two blocks at its main data center since converting to Nutanix, with plans for more as utilization grows.

"We're trying to expand ahead of the business -- watching utilization ... and spinning up new services," he said.

Changes to the physical infrastructure are simplified, eschewing a host of third-party tools to manage the migration. This simplified Francis Drilling's plan for a move to a new facility.

"We can copy everything over to the back-up OmniCube, move the boxes and bring production back up on them," Schaaf said.

Power and cooling

Implementing integrated systems may require changes to the facility infrastructure. As with refresh cycles, adopters tend to plan the new infrastructure at an inflection point for capacity.

"We were at the limit for our cage [at the colocation facility], so we tied in the upgrade with a move to a higher power density floor in the data center," McGlade said. With the added floor space and upgraded power cables, there was no worry about the vBlocks' resource usage.

Francis Drilling, before changing facilities, ran the SimpliVity boxes alongside its old IBM blade server infrastructure while working to transfer all its data. With both types of systems powered up, the temperature in the data center didn't change, Schaaf said.

One for all?

Converged and hyper-converged architectures aren't the right fit for all organizations. Larger enterprises with dedicated and skilled IT teams won't see the same management benefits as others, TBR's Perry said. This is true of some smaller companies as well.

"I'm very hands-on when it comes to anything technical," said Drew Green, IT director at Thomas, Judy & Tucker, who decided to upgrade the N.C.-based accounting firm's infrastructure with a StorTrends high-performance SAN rather than explore hyper-converged options.

Converged and hyper-converged infrastructure is more expensive upfront and requires a compressed refresh model compared to the staggered approach of server, storage and networking purchases. Converged architecture adopters are frequently at an inflection point of capacity, so CI boxes don't push out fully functional, new hardware. Users say the ongoing costs -- maintenance, integration, support and scaling -- make it less expensive than managing a mix of separate components.

Meredith Courtemanche is the site editor for SearchDataCenter. Follow @DataCenterTT for news and tips on data center IT and facilities.

Dig Deeper on Converged infrastructure (CI)