Netuitive 5.0: Self-Learning Service Management Meets Cloud Infrastructures

Diagnosing application performance problems is never easy for IT. Application engineers talk one language, systems engineer speak a different language. Reconciling those views in most organizations is the Holy Grail for most IT departments.Netuitive, Inc would like to be the one to deliver on that promise. The application performance management company announced today that its newest release, Netuitive 5.0, would combine its system performance management console,  System Intelligence (SI), with its application performance management console, Service Analyzer (SA).

With version 5.0, application engineers and system engineers would finally have a single platform for diagnosing application problems end-to-end. An engineer could click on a cloud service offering, such as online storefront, in the 5.0 console and view the sub-services that might make up that complete cloud service offering, such as authentication, catalog look-up, inventory and credit card authorization, even if some of these are enabled by third-party providers. The engineer could then diagnose the cause behind a failure or slow down in cloud service.

The other major addition in the new release is a backend database, the Performance Management Database (PMDB) for capacity management. The PMDB integrates raw performance data collected by existing systems monitoring tools, such as from BMC, CA, IBM, HP, Microsoft and VMware. By drawing on the performance data, organizations can analyze resource utilization from hundreds of different angles, such as identifying the most under/over-utilized servers, identifying factors driving the workload for a group of servers or analyzing resource usage by application, region, owner, business unit or service. Users can also export data to third-party enterprise reporting and analysis tools for long-range capacity planning.

Cloud computing shook up IT in fundamental ways. Management of a virtualized IT infrastructure requires being able to adapt dynamically to changing conditions. Simply setting threshold on CPU utilization isn't enough when multiple servers maybe sharing that processor. Yet all too often application performance management platforms have been static forcing IT managers to do just that.  

"You use your experience to decide. As an atomic system, you take some measurements and it's easy to setup a threshold, but in a virtualized environment, it becomes a more complex system." says Henry Mayorga, director of infrastructure at Baron Funds, an investment firm and a Netuitive customer. "Now hardware is abstracted and software looks like hardware. Figuring out what those thresholds should be is much more complex. If you're evaluating a host CPU but don't know if being shared between one, two or three services, it's very difficult to do." Analysts have reported something similar. "Cloud computing is quickly gaining traction in the marketplace, yet private, cloud-based infrastructures are far more dynamic than anything a data center manager could have envisioned even five years ago," comments Dennis Drogseth, Vice President with Enterprise Management Associates.To address these challenges, Netuitive uses self-learning and continuously adaptive software. The company claims that software is the culmination of nine patented technologies and 20 years of research. Netuitive 5.0 monitors the network and uses automated mathematics and analysis to understand normal system behavior across IT silos, isolate root causes of service issues and forecast degradations before they impact performance.

So does it work? Mayorga thinks so. Last year after a three day weekend, Mayorga and his team came back to the office only to find Exchange service slowed to a crawl. Nothing had changed in the network, yet Exchange was simply not responding.  Normal diagnostics showed nothing, but clearly something was up. That's when he remembered the Netuitive software he had installed a few weeks previously. Mayorga went to the Netuitive console "and I saw a big red circle on the ESX host," he recalls, "I drilled down on the host and it showed that I/O was way off the scale and responding 10 times slower."

The problem? A woman had come in over the weekend and went through the graphics files on the systems backup looking for graphics for a new brochure. The Compellant subsystem, seeing the traffic activity, registered that these files were active and migrated them to tier one storage and moved the Exchange files to a lower performing tier since they weren't in use. When Mayorga and his team resumed work, Exchange was hitting a slower tier, slowing down the I/O system.  "That made a convert out of me," he says.

Yet for all its promise, Mayorga warns that IT professionals should approach 5.0 with caution. "Netuitive is a fantastic engine for figuring out trends, but you must feed it the right data and getting data into Netuitive world is challenging," he says. Netuitive integrates with specific platforms, such as Tivoli and HP Openview, but that's not sufficient for many organizations. "They have all the hooks up for big business, but they don't have the hooks for small businesses I can pull data from VMCenter, but really want data form filers (Compellant), Cisco switches, and i.Link interfaces from my HP servers. Getting them into Netuitive is not trivial."

As such, while Mayorga buys into the Netuitive vision of service management, he's not there in a practical sense. Getting there and expanding those options will be critical for performance management vendors. A standard, such as IF-MAP could help a lot on this score. Netuitive president, Nicola Sanna, president and CEO of Netuitive, acknowledged as much, but said that he'd wanted to wait to see if there would be greater industry adoption of the protocol. Whether through IF-MAP or some other mechanism, until Netuitive can integrate with more data sources attracting widespread enterprise, IT deployments will remain a challenge.  Pricing for Netuitive 5.0 is per managed element. Each server runs $1,000 to manage, $12,500 per socket for VMware, and $20,000 for an application. The product is currently available.