Optimising AI Performance with Juniper Networks and AMD

A leader in secure, AI-Native networking, Juniper Networks has introduced an industry-first Ops4AI Lab to maximise AI workloads using open and flexible infrastructure that is easy to manage. 

The initiative is designed to accelerate the deployment and optimisation of AI infrastructure across the data centre sector. Additionally, the company is releasing advanced software enhancements and a collaborative ecosystem to optimise the performance and management of AI workloads over Ethernet. 

Accelerated time-to-value is assured by the company with Networking for AI configurations, alongside AMD, Broadcom, Intel and Nvidia technology. Juniper Networks is collaborating closely with these ecosystem partners to enable the best AI performance via the most flexible and easiest-to-manage data centre infrastructures.

The news comes in the midst of significant AI expenditure, with Gartner predicting a US$5.26tn global IT spend in 2024 - with businesses harnessing the power of generative AI (Gen AI) to remain competitive.

Optimising data centre operations with AI

Juniper Validated Designs (JVD) are implementation documents that offer new customers confidence that their chosen solution is well-tested and repeatable, resulting in faster time to successful deployment. They aim to provide assurance and are tested in best practice designs based on specific platforms and software versions.

The company has released the first blueprint specifically for AI data centres that is built on Nvidia A100 and H100 compute, in addition to storage from Juniper’s ecosystem partners. As a result, the new Ops4AI JVD complements its existing JVDs for automated and secure data centres, which includes data centre automation and Juniper’s range of data centre security solutions.

As a key part of Juniper Networks’ AI-Native Networking Platform, the existing networking for AI solution consists of a spine-leaf data centre architecture with a foundation of AI-optimised 400G and 800G QFX Series Switches and PTX Series Routers. It’s secured via high-performance firewalls and managed via Juniper Apstra, an intent-based networking software.

“Best-of-breed always wins out, and the same will be true for compute, storage, networking and operations in AI data centres. Juniper has made a significant investment in the Ops4AI lab, JVDs and a new promotional programme to enable our customers and partners to have maximum choice, flexibility and stability in how they build a complete Gen AI solution,” comments Praveen Jain, SVP and GM of AI and Data Center at Juniper Networks. 

“There has never been a better time to build high-performance, low-latency, multivendor AI data centre solutions that are simple, fast and economical to deploy and operate.”

Currently, data centres serve as the infrastructure backbone to the growing AI ecosystem. However, the technology is using an extreme amount of electricity and energy globally, putting pressure on the industry to be more sustainable. As a result, businesses within the sector are needing to be more innovative to meet the demands of power-intensive AI applications safely.

Streamlining AI performance

Openness and collaboration are essential to Juniper’s networking mission. The company posits a way to move AI data centres from their current early adopter stage to effective mass market deployments. End-to-end operations for multivendor AI data centre infrastructure has been difficult, leading to vertically integrated AI data centre solutions that are vendor-locked and lead-time challenged. 

Automated operations with switching, routing, storage and compute solutions from leading vendors, as well as new Juniper Validated Designs (JVDs), accelerate the time-to-value in deploying AI clusters.

Juniper Apstra provides key Ops4AI capabilities, such as intent-based networking, multivendor switch management, workload awareness, AIOps proactive actions and a Gen AI conversational interface. With Juniper’s full Networking for AI solution, customers and partners are able to lower AI training Job Completion Times, reduce latency during inferencing and increase GPU utilisation - all while decreasing deployment times by up to 85% and reducing operations costs by up to 90% in some cases.

Whilst AI has changed the game for data centres worldwide, the technology is also calling for higher performance and scalable infrastructure. To simplify AI clusters and maximise network performance further, Juniper has added new Ops4AI software enhancements that together offer unique value for customers.

“Truly pervasive and performant AI infrastructure relies on standards-based technologies, open source software and industry wide collaboration through organisations such as the Ultra Ethernet Consortium,” says Steve Scott, Corporate Fellow, Network and Systems Architecture at AMD. “AMD, Juniper and our partners across the ecosystem, bring together extensive experience in creating and deploying high-performance, low-latency networking solutions to deliver AI innovation.”

******

Make sure you check out the latest edition of Technology Magazine and also sign up to our global conference series - Tech & AI LIVE 2024

******

Technology Magazine is a BizClik brand