Cisco Systems Inc.’s AppDynamics subsidiary is broadening its monitoring scope and folding machine learning into its technology in a campaign aimed at bringing more automation to complex multicloud environments.

The company, which Cisco swooped in to acquire two years ago just hours before AppDynamics’ planned initial public offering, monitors systems from an application perspective, detecting slowdowns and outages at the user experience level and digging down to identify root causes. The enhancements leverage technology contributed by Perspica Inc., another Cisco acquisition, which uses machine learning to analyze streams of real-time data.

AppDynamics said today’s announcements are part of a long-term initiative to build what it calls a “Central Nervous System for IT” that spans applications, infrastructure and networks. The company is tackling the increasingly vexing problem of how to sift through and make sense of data streaming in from network equipment, databases, systems monitors and other sources, a task that’s become more complex as companies have begun using multiple clouds.

The three-stage Central Nervous System initiative covers visibility, insight and action, said Matt Chotin, senior director of developer initiatives at AppDynamics. “We’ve already offered visibility; this is the core of the insight pillar,” he said. The action component, which is still under development, will involve automating problem resolution.

Cognition Engine can ingest, process and analyze millions of records per second, using machine learning to improve its understanding of how those metrics correlate with each other. The software isolates metrics that deviate from what it perceives to be the norm and presents the top suspects of the root cause of any issue affecting application availability or performance. “Instead of clicking through screens to find the root causes, the engine is elevating them and suggesting what they might be,” Chotin said.

The machine learning component is unsupervised, meaning that human intervention isn’t required to define normal conditions, Chotin said. “We collect the data, feed it into the model and over time it understands what normal is,” he said. “You don’t really need to train it. You do need to understand what your business thresholds are for deviation,” which is a statistical measure of tolerance for abnormal behavior.

Data collection is done with agents that are installed and connected to runtime module such as Java applications or Docker containers.

The company is also adding a serverless agent for Amazon Web Services Inc.’s Lambda function-as-a-service, the first of what the company said will be a family of special-purpose agents that detect the impact of serverless functions on applications. Serverless computing presents a new set of challenges to information technology organizations because functions can be launched on the fly and shut down just as quickly. This ephemeral nature of serverless applications defies traditional system monitoring technology.

AppDynamics said the new Serverless Agent for AWS Lambda creates a full application topology that understands how applications are built and how different components, including serverless functions, impact user experience. It uses a lightweight agent that can scale without requiring significant network or system overhead, the company said. While acknowledging that “serverless is more aspirational than deployed at this time,” Chotin said the company wants to get out front of what is shaping up to be a major shift in the way organizations build and deploy applications.

Cognition Engine is an enhancement to the base AppDynamics platform, so pricing won’t change. AppDynamics typically prices per deployed agent, but doesn’t publish specifics.

Image: Pixabay