Cloud Application Management Complexity A Growing Challenge

IT organizations are already using cloud in highly sophisticated ways in spite of the added level of complexity it introduces in managing applications. That was one of the surprise findings from Enterprise Management Associates' "Radar for Application Performance Management for Cloud Services: Q1 2012," according to the report's author, Julie Craig, research director, application management, at EMA.

"As application management becomes much more complex as they turn to cloud, [IT organizations are] using cloud in ways that surprised me in terms of sophistication,'' Craig says. For example, almost half of the companies surveyed are running tiered transactions/services spanning both cloud and on-premise, according to the report, while 35% have either integrated or are in the process of integrating multiple software-as-a-service (SaaS) applications.

"The research showed quite a large number of companies that are actually already running transactions that span multiple SaaS services, and I didn't expect them to be this far along,'' Craig notes. "They are very sophisticated in terms of deployment but struggling as an industry with finding APM products that can deal with this kind of complexity." Her research found, for example, that many public cloud providers do not yet offer monitoring agents or APIs, which is hampering vendors' ability to build capabilities management into APM products.

Additionally, the majority of midsize to large businesses have already embraced private cloud as a viable delivery option for business-critical applications, the report finds, as some 66% of companies are either already using infrastructure as a service (IaaS) or planning to do so within the next year. Among other findings are that 44% of companies are already utilizing at least one SaaS service, with another 33% planning to do so in the next year.

There are a couple of key challenges in terms of managing cloud applications, says Craig. "One is that you have to build an end-to-end picture of the transaction or application. Unless you have that [visibility], you don't know where to start in terms of actually solving application performance problems." But the challenge is being able to drill down and understand as many as 30 to 40 components supporting the application to determine the origin of a slowdown, for example. "You have to be able to see the application from start to finish in context with the technology that's supporting it,'' she says. "IT organizations are starting to experience this problem. In many cases they're trying to manage these very complex environments with manual kinds of activities, and they see the need to get away from that because it's taking a tremendous amount of time." It's also very expensive to manage these complex environments with people alone, she adds.

In spite of the struggles APM vendors are having, Craig was also surprised by "how fast vendors are evolving products to address" the complexity during the timeframe she conducted her research. "There are constant updates and enhancements to cloud APM solutions," she says, "so it's a very rapidly evolving market." APM product vendors covered in the report include OpTier, AppDynamics, AppFirst, Aternity, CA, Compuware, Correlsense, eG Innovations, HP, IBM, INETCO, Nastel, Netuitive, New Relic, OPNET, Quest, SolarWinds and Splunk. More than 40 users of the products were interviewed for the report.

But even with the upgrades and enhancements, Craig doesn't see the level of APM complexity diminishing anytime soon. "It's still an evolving market,'' she says. "The task of managing applications is probably the most challenging of any of the enterprise management disciplines. Application management relies on assimilation of metrics from across the entire application execution ecosystem."

Her vision is to see APM become "increasingly automated," although full automation capabilities will likely take five to 10 years to evolve. Flow-based network analytics will likely be a powerful force in enabling full automation of APM systems, Craig believes. Network analytics leveraging such information to track and model end-to-end application execution will be the answer to enabling greater levels of automation in the future, she maintains. "I believe the network will ultimately provide the unifying information necessary to automate the process of application performance management,'' Craig says, adding that within the next two to four years, "APM vendors are going to turn to the network in very creative ways to build this end-to-end view of application execution."

Learn more about Research: 2012 State of Cloud Computing by subscribing to Network Computing Pro Reports (free, registration required).