Unified performance management goes beyond APM

Unified performance monitoring and management promises fewer, lighter, more effective IT tools to manage servers, applications and more.

ORLANDO, Fla. -- The lack of interoperability among traditional performance monitoring tools means data centers must use -- and pay for -- multiple tools. But movement toward unified performance monitoring and management could change all that. 

Tool vendors want to move toward a monitoring environment that uses fewer, more effective lightweight tools, according to Jonah Kowall, research VP, who presented during a unified management session here at Gartner's IT Infrastructure and Operations Management Summit this week.

Here are the limitations of traditional performance monitoring tools and the potential benefits of next-generation unified monitoring.

Traditional performance monitoring tools fall short

Every organization uses software tools to measure performance. This can include application performance monitoring (APM), network performance management (NPM), server performance management, storage performance management and other data center tracking tools. But traditional products present a variety of problems that are difficult for IT designers and architects to overcome.

APM, for instance, is an essential indicator of the user experience and the quality of that application's service to the enterprise. But traditional APM tools scattered across a data center may not provide the information IT leaders need to gauge availability.

Often it's the sheer number of tools that poses the most trouble. Many businesses rely on silos for core tasks like server, network, storage and even virtualization management. Since most tools are based on proprietary software, they typically do not integrate or share data -- limiting the view between silos.

The tool count multiplies as each functional area justifies and introduces new software.

Kowall polled the Gartner conference attendees and found that 40% of respondents rely on six to 10 different tools to monitor and manage performance in their environment.

Data gathering methods pose another problem. Agents are the most common way to query and relay data from monitored elements, but agents can affect performance, and older agents can become a potential vector for attack. Both are frequently overlooked since data center architectures lack a cohesive plan to accommodate or maintain monitoring agents. Still, there is an element of vendor lock-in, since many tool vendors continue to push agent-based data collection and delivery.

But these complaints do not address the changing reality within modern IT organizations. The silos that kept organizations apart are no longer static; applications are increasingly interdependent and dynamic. For example, applications may depend on databases or on other applications, and may need to scale computing as user demands change. As a consequence, provisioning an application involves greater interdependency between networks, storage, servers and so on.

Application monitoring must do a better job of gathering and reporting data across a broader scope of data center functional areas -- yet impose lighter requirements and operate more efficiently within the environment.

Unified management gains ground

The idea of unified performance monitoring and management has gained traction among tool vendors and users. The goal is to provide a simple and cost-effective tool to operate at superior levels of autonomy across functional areas without the pitfalls of agents.

Unified performance monitoring utilizes the rich range of application programming interfaces (API) now available to query systems to gather and report a growing range of information, Kowall said. A tool that exists at such a high level can also gain insight into connected systems, for better discovery and topology analysis. Some unified monitoring tools can also perform log file analytics to correlate events and aid troubleshooting across dozens --even hundreds --of devices in complex environments.

The results are significant. Consider that a virtual machine can be spun up in a matter of minutes, but monitoring that new VM often requires the manual installation of agents, along with costly licensing and time-consuming configuration to add that VM to the range of monitored machines. An ideal expression of unified monitoring is to discover and respond to the new VM automatically using existing APIs that need no further software installation and little (if any) setup.

The list of unified monitoring vendors is growing -- particularly as some well-established vendors embrace lightweight, API-driven monitoring along with log file analytics capabilities. Kowall points to vendors including AccelOps Inc., CA Technologies Inc., Centerity Systems Inc., GroundWork Inc., Ipswitch Inc., Kaseya, ManageEngine, ScienceLogic Inc., SolarWinds Inc. and Zenoss Inc. Vendors adopting log file analytics include HP, IBM, Elasticsearch, VMware (Log Insight), Splunk Inc. and SumoLogic.

Unified performance management growing pains

Although the benefits of unified monitoring and management are growing, the technology isn't quite ready to displace every other APM or NPM tool just yet. But unified management can certainly lower the total tool count, greatly reducing the burden and risk of older legacy tools, while complementing well-selected APM and NPM products, Kowall said.

For example, tools designed to measure the end user experience and provide early warnings of possible application performance problems may continue to rely on tailored, agent-driven tools. This is because agents tolerate disruptions in connectivity, quietly gathering data to share with a centralized collection server when connectivity is available.

API-based monitoring tools are not so forgiving. Organizations may also continue to use other tools for specific tasks like monitoring key business transactions or complex troubleshooting, performance tuning and capacity planning.

An ongoing challenge for unified tools is that reliance on APIs requires other software vendors (which provide the APIs in the first place) to do more work in their product development and testing. The landscape is constantly improving, but without standards there is no assurance that a suitable, uniform API suite will become available -- some monitoring may simply not be available or consistent across heterogeneous platforms.

The promise and pitfalls for data center organizations is clear. Introduce unified management in well-planned proof-of-principle projects, perform copious testing, and be sensitive to the various product roadmaps that might affect API availability into the future. Only then can you roll out unified performance management in your data center.

Dig Deeper on Real-Time Performance Monitoring and Management