Arista Networks has announced its new CloudVision Universal Network Observability (CV UNO) product, promising to deliver unified visibility across network, compute and application layers.

With the growth of hybrid cloud environments, IT teams often struggle to pinpoint the root causes of performance issues spanning different domains. With CV UNO, Arista is hoping to solve this problem by merging various data sources into a single view.

Arista's new platform aims to provide a deep level of rich contextual data and insights about network and application dependencies. It pulls in real-time state data from the network as well as other sources like identity management and virtualization systems.

This data is stored in a centralized data lake comprising various database technologies. Machine learning (ML) algorithms then analyze the data lake, identifying patterns and correlations to enable risk analysis, impact analysis and root cause identification.

“We've always had with CloudVision the ability to capture network state,” Douglas Gourlay, VP and GM of software at Arista Networks, told SDxCentral. “What we didn't have was the ability to talk to adjacent systems of record and build this dynamic real-time updating application and network and dependency graph.”

A key aspect of CV UNO is integrating data from a variety of systems to build a comprehensive view of the network.

Gourlay explained that CV UNO is connected to multiple systems of record including IP address management (IPAM), configuration management databases (CMDBs) and domain name system (DNS), as well as vendor technologies such as ServiceNow and VMware.

“We ingest that data, we do a series of unions and joins in a graph database that then allows us to create context,” he said.

Arista joins all the database system information together in a data lake. Gourlay said the data lake is a series of different multimodal databases that are abstracted with a common API surface.

The complexities of the different databases are abstracted away so data can flow in and be queried using a common language. This allows the creation of a dynamic and real-time application and network dependency map that gives visibility into relationships and dependencies across systems.

Gourlay noted that this context enables CV UNO to do queries about risk management, impact analysis and root cause analysis.

In addition to the integrated data, Arista is leveraging artificial intelligence (AI) and ML techniques to provide smarter insights from the data.

Gourlay explained that Arista is using both supervised and unsupervised ML models. The supervised learning models are used to help understand repeating patterns that indicate application performance issues. Unsupervised learning is used to automatically tag entities like applications, hosts and users with a very high precision rate.

The models help infer things like the impact of a 1.5% packet loss on different application types. Overall, AI and ML are used to analyze patterns in the centralized data lake and provide recommendations, predictions and insights to network operations teams.

Gourlay emphasized that a key goal of CV UNO is to improve network operations, reducing the need for manual troubleshooting and firefighting.

“Our buyer for this, to be very blunt, is a network operations or network engineering team that wants to more effectively manage risk, have more insight into change management and who wants to be able to have a clearer understanding of root cause analysis when unintended consequences happen,” he said.