How Mt. Sinai's IT Team Made Virtualization Easier

CIO — In May, 2009, almost a year after Microsoft released the hypervisor it planned to make ubiquitous and three years after VMware's virtualization infrastructure sparked a data center revolution, more than half of the IT managers polled at one conference said there were still not enough tools available to help them realize more benefits than problems with virtual servers.

Six years ago, when Prateek Dwivedi, VP and CIO at Mount Sinai Hospital, started virtualizing servers for the first time, there were so few tools available and so few people skilled in virtualization that the whole process was like feeling around in the dark, says Dwivedi.

"Now it's practically easy," he says. "I was happily shocked to see that was possible. Almost all of our applications can be virtualized except a couple with hardware dependencies, and our capacity planning showed that we could consolidate 135 [physical] servers down to seven."

Key to his virtualization maintenance and automation is up.Time Software's suite, used to monitor SLAs and check the health of both physical and virtual servers.

Tools Landscape Matures

VMware, Microsoft, Citrix, Hewlett-Packard and almost every other major vendor has shipped a suite of management tools designed to tighten security, automate performance management, monitor interactions among virtual servers and other critical functions.

However, almost none have yet rolled out software that's able to take a macro view of both a data center and a virtual infrastructure, monitoring and controlling physical and virtual servers in a way that makes it easier for data-center managers to maintain service levels and troubleshoot problems as they come up, according to Andi Mann, analyst at Enterprise Management Associates (EMA).

[ For timely virtualization news and expert advice on strategy, see CIO.com's Virtualization Drilldown section. ]

Systems and software management vendor uptime software this week introduced a new version of its product, up.time 5, that is integrated with VMware's vCenter Orchestrator, a process-automation tool that shipped in April with VMware's vSphere 4.0 suite of infrastructure management tools.

Orchestrator is designed to create workflows within VMware infrastructures, invoking third-party applications as necessary to accomplish certain tasks. But up.time 5's automation expands beyond VMware to include physical servers and non-Windows servers, helping users create "incidence avoidance" plans in which specific warning signs will trigger the data center to shift the applications from one VM to another, for example, says uptime CTO Alex Bewley.

"Where uptime is good is it does both the virtual and physical infrastructures for a good range of stuff," Mann says. "They're not going to replace the big four—BMC, HP, Tivoli and CA. They do service level monitoring and management, not system and app provisioning and patch management and that sort of stuff. They're a good complement, though."

Mount Sinai uses uptime to monitor its own service levels, verify the SLA data that its data-center management service provides, and to help with capacity planning and migration on a virtualization project that includes both Hyper-V and VMware-based servers as well as a range of Unix servers and specialized applications.

The Automation Advantage

"Runbook automation is really a big thing, for us," Dwivedi says." We've got two data centers; one with mission-critical apps is managed by a service provider using HP OpenView, which they use to report on how well they're meeting their SLAs. They tend to be very infrastructure focused; it doesn't tell you much about how the applications are performing."

Dwivedi has been testing new features in up.time 5, along with VMware's Orchestrator process-automation software, to monitor application performance from the end-user's point of view and identify where reported problems or slowdowns are real, and where they're the result of how clinicians use them.

One physician who complained that his PC and the network were slow turned out to have seen a patient who brought in a DVD filled with "multiple multi-gigabyte files" of diagnostic imagery from another hospital. "He said he had to look at it right then, but he didn't know putting it on the network—there's never enough capacity for that," Dwivedi says.

Run-book automation (RBA) is the ability to create, execute and monitor workflows—a series of specific actions various applications or servers will take under predetermined conditions—to provision additional servers when the load is heavy, schedule a live backup when the power gets iffy, or other ways to make sure the data and servers continue to act as desired.

"We don't run into situations where we have a surge of servers we have to put online to cover demand, but where a server is having some issues, being able to bring another one up and move over to it without losing anything, that's critical to us," Dwivedi says. "I'm looking forward to integrating that with the monitoring and capacity planning."

Service level monitoring and management are important, especially in virtualization rollouts, but process automation can be a huge contributor to a data center's performance levels, Mann says.

"Our data shows enterprises with IT process automation are achieving better outcomes on their SLAs, using less staff, having shorter mean-time to repair—a lot of improvements in how the data center runs," he says. "The IT automation discipline has gotten a lot more important over the last 12 months or so; that's why HP acquired OpsWare and CA acquired Optinuity and IBM came out with a Tivoli Service Automation Manager. It's becoming very important. It's interesting that up.Time is getting in there."

Do you Tweet? Follow everything from CIO.com on Twitter @CIOonline.