In-Depth

Netuitive Pushing Adaptive Learning for Virtualization Management

As data centers undergo increasing virtualization, a more holistic approach to managing VMs across organizational boundaries may be needed.

• By Tom Valovic
• 12/12/2008

One of these is Netuitive, based in Reston, Va., and headed up by CEO Nicola Sanna. Netuitive has been in the BSM and performance management business since 2002 and currently has over 200 customers including Cigna, Verizon, NTT, the US Army and MetLife. The company's secret sauce is a patented algorithm that its performance management software uses to provide continuously adaptive, self-learning capability -- the territory of artificial intelligence.

More recently, Netuitive has entered the virtualization space. The ramp-up has taken place fairly quickly, and currently five percent to 10 percent of the customer base is tied to virtualization offerings.

The company's first virtualization product was Netuitive SI for VMware, available since March of last year. Intended primarily for server virtualization, the software self-learns the behavior of each virtual machine (VM) and host running in a VMware ESX environment. It then adaptively correlates VM to host performance, according to Daniel Heimlich, vice president of marketing, without using any manual rules, scripts or thresholds.

"When you're talking about a virtualization environment which is much more interdependent, dynamic and fluid, you're talking about exponential levels of complexity, in terms of trying to understand all these different performance behaviors based on time of day or day of week" Heimlich says. "What [the product] is essentially doing is automating all that admin effort."

Adaptive learning is based on an algorithm that performs sophisticated mathematical trends analysis and multivariable correlation. Through this process, the software first learns both normal and abnormal system behavior. From that baseline, it then is able to forecast and interpret deviations to create alerts and provide various levels of diagnosis.

"The core algorithm is very data-agnostic which is part of its beauty" says Dennis Drogseth, an analyst with EMA Associates. "Its origin was pricing oil based on macroeconomic conditions…. [The software] understands relative weight and impact [of variables] to an outcome. This is what makes it so valuable."

Using this approach, Netuitive says SI for VMware can provide degradation forecasts up to two hours in advance of the event. In addition, it isolates root causes and provides necessary corrective actions. It also identifies which VMs might be choking the performance of an ESX server.

Heimlich says that a major part of the company's marketing approach relates to the notion that virtualization management needs to be done on a horizontal and not a siloed basis. In the current IT environment, management of various systems, such as databases, servers and desktop devices, breaks along separate organizational lines. The company contends that when it comes to managing VMs, the danger for IT shops lies in creating yet another system silo.

Instead, Heimlich argues that virtualization management routinely needs to be handled from a system-wide perspective.

"The server guys know how the servers are running and the database guys know how the databases are running…but [neither] understands how the relationship of the two provides an overall experience to the customer," he says.

In a VMware environment, Netuitive SI for VMware correlates VM performance data that's available on VMware's management platform, called vCenter (formerly Virtual Center). Since Netuitive is a VMware partner, the product has been designed using APIs to access the performance data that vCenter gathers.

Future releases of the product will take the process a step further by correlating it with end user performance data being generated by another VMware management product called AppSpeed. Netuitive is also planning to adapt the product to work with Microsoft's core hypervisor Hyper-V and Citrix's XenServer.

In June of this year, the company came out with another virtualization product, an enhancement to the Netuitive Service Analyzer, which the company describes as the "industry's first self-learning correlation software for managing performance of enterprise applications across mixed physical and virtual environments." As with its other products, adaptive self-learning is also the centerpiece of its capabilities.