Juniper Networks Inc. today introduced a new artificial intelligence tool, Marvis Minis, that can simulate the impact of a network change to determine if it may lead to technical issues.

The company is rolling out the software as part of a product portfolio refresh that also extends to its hardware lineup. Juniper today debuted new data center routers and a switch that provide 800GE connectivity, or the ability to process up to 800 gigabits of data per second per port. The new routers use the company’s internally developed Express 5 chip, while the switch ships with a five-nanometer processor that Broadcom Inc. began mass-producing last March.

Marvis Minis is rolling out for Juniper’s Marvis platform, which uses AI to help administrators find the root cause of network malfunctions. The software automatically generates troubleshooting suggestions. Additionally, it provides a conversational interface that administrators can use to ask for pointers about tasks such as setting up a router.

According to Juniper, its newly debuted Marvis Minis tool can automatically learn how a network is configured. It then simulates connections from users, those users’ devices and the app running on the devices. This simulated traffic makes it possible to surface network malfunctions that would otherwise only be brought to administrators’ attention after causing technical issues for workers.

After it spots a network issue, the tool sends the incident data it collects to another feature of Marvis called Marvis Actions. The latter capability automatically generates troubleshooting recommendations for each malfunction. Additionally, it can fix some types of technical problems without manual input.

Juniper the fact Marvis Minis is integrated into its Marvis network automation platform is a benefit in itself. According to the company, that integration removes the need to install additional, external software components in a network to collect troubleshooting data. Reducing the number of software tools that administrators must learn and maintain can ease their day-to-day work.

Marvis Minis is rolling alongside a second new AI capability called Marvis VNA for Data Center. According to Juniper, it will make it easier for administrators to access information about a data center network’s cabling, configuration and other technical details. The company says that the capability works both with its own network equipment and hardware from competitors.

“AI is the biggest technology inflection point since the internet itself, and its ongoing impact on networking cannot be understated,” said Juniper Chief Executive Rami Rahim. “By extending AIOps from the end user all the way to the application, and across every network domain in between, we are taking a big step toward making network outages, trouble tickets and application downtime things of the past.”

Juniper detailed the software enhancements alongside an update to its AI Data Center offering. It’s a product bundle that includes Marvis and the company’s Apstra network management platform, as well as its PTX routers and QFX switches. The offering is designed to ease the task of building networks optimized for artificial intelligence applications.

The AI Data Center implements a so-called spine-leaf architecture. This is a type of network design that helps reduce hops, a term for the number of network devices through which a packet must pass while traveling between two servers. Reducing the number of hops that traffic must go through can speed up connections and thereby boost AI workloads’ performance.

The first component of the AI Data Center for which Juniper debuted an update today is the Apstra network management platform. According to the company, the software can now more quickly process traffic from AI applications that is zipping through Ethernet infrastructure. Juniper is also promising efficiency improvements.

The PTX router lineup that powers the AI Data Center offering is receiving an update as well. Juniper is introducing new routers and line cards based on Express 5, an internally developed network processor that it developed in 2021. The chip is made using a seven-nanometer manufacturing process and can manage up to 28.8 terabits of traffic per second.

Juniper’s QFX switch family, in turn, is being expanded with the addition of a new device that feature Broadcom’s Tomahawk 5 chip. It’s a five-nanometer chip series that entered mass production last year and can process up to 51.2 terabits of network traffic per second. Broadcom has equipped the Tomahawk 5 with a suite of performance optimization features, dubbed Cognitive Routing, that can route traffic around congested network links to improve connection speeds.

Photo: LPS.1/Wikimedia