Juniper Networks’ Data Center team recently spent time with 12 industry luminaries and practitioners at Cloud Field Day 20 (CFD 20), demonstrating how businesses can “seize the moment” to deploy artificial intelligence (AI) data center infrastructure to take advantage of recent advancements in AI and machine learning (ML). It was an exciting event with fantastic engagement from the CFD delegates: “Juniper has built some amazing technology”.
Juniper presented five sessions at CFD 20, diving into detail on AI data center networking, or what we call, “Networking for AI”:
1. AI transformation is here. Is your data center up to the task? Mansour Karam, GVP Data Center Products, discussed how the evolution of the internet over the last 20 years led Juniper Apstra to an operations-first approach to transform the data center network. This motivated Mansour and team to invent intent-based networking (IBN), which has revolutionized the way customers manage the entire data center lifecycle, from Day 0/Design to Day 1/Deployment through Day 2+/Operations. But AI is now all the rage and the combination of IBN and AIOps in the data center will be seismic.
Juniper’s AI-Native Networking Platform is optimized for the connectivity, data, volume, and speed requirements of mission-critical AI workloads. Our AI data center solutions offer the fastest and most flexible way to deploy reliable high-performing AI training, inference, and storage clusters, and the simplest to operate with limited IT resources. With the data center at the heart of the most exciting area in all of tech we are committed to making yours a success.
2. In the rapidly evolving landscape of AI data centers, misconceptions abound, often clouding the decision-making process for IT professionals. Praful Lalchandani embarked on a myth-busting journey to unveil the realities of AI data centers. His session debunks common myths and showcases how Juniper Networks’ cutting-edge networking solutions can transform and optimize your data center infrastructure.
Can Ethernet handle the demands of AI training? Is packet spraying required to maximize performance in backend GPU fabrics? Is 1:1 non-blocking a must for maximum performance? Is lossless networking always required in the backend fabric? Do we need more acronyms in the AI world? Tune in to the session to discover the answers to these important questions.
3. Private data center as easy as cloud. Many of the new network engineers today are not “networking by birth”, they are “cloud by birth”, meaning they’re more likely to know HashiCorp® Terraform® instead of Cisco, Arista or Juniper CLI. Nick Davey, Director, Product Management and Raj Subrahmanian, Product Manager, showed us how to achieve public cloud-like service consumption in your on-prem AI data center with Juniper Apstra and Terraform. Juniper Apstra and the Terraform Provider for Juniper Apstra fit traditionally complex network services, like EVPN, neatly into predefined application automation. The intent-based nature of Juniper Apstra aligns perfectly with Terraform’s declarative approach. Network teams can self-serve network services in a familiar way, providing seamless deployments across any infrastructure for new AI/ML workloads.
4. Struggling with where to start with your on-prem AI training cluster? Jay Wilson, Data Center Architect explains that, while Juniper Apstra is excellent for running traditional data centers that support standard applications and workloads, it’s also ideal for new AI training, management, storage, and inference clusters. See how Apstra does the Day 0/1/2 heavy lifting for you with intent-based automation. Jim Czuprynski saw “Another bunch of neat features of Juniper Networks’ Apstra offering. . . This is what #CFD20 is all about: put yer money where yer mouth is, dear vendor.”
Juniper validated designs (JVDs) are rigorously pre-tested to make sure deployments are relatively pain-free Chris Reed remarked, “woah, the @JuniperNetworks Validated AI designs aren’t just PDFs, they have Apstra templates that you can deploy to match.”
5. To maximize throughput and minimize packet loss, Ethernet uses the DCQCN congestion management protocol, but DCQCN introduces significant operational complexity for human operators. Raj Subrahmanian and Vikram Singh, Sr. Product Manager, walked through how Juniper Apstra handles this new challenge in stride, automatically optimizing throughput vs. packet loss.
Seize the AI moment
We capped off the day with an exclusive, off-camera tour of the Juniper AI Lab in Sunnyvale where we showed our full stack AI data center setups for customer proof-of-concepts. But not too exclusive: here’s a quick tour of the lab. Customers often bring-their-own-model (BYOM) or bring-their-own-data (BYOD). We poke and prod and test models across multiple real-world scenarios, auto-tuning the infrastructure to drive optimal results.
The sessions are all available online and we invite you to watch them to see what the delegates and over 1,500 live viewers saw livestreamed from CFD20.
Juniper would love the opportunity to speak with you and discuss how we can help you seize your AI moment. Feel free to reach out to your local Juniper representative or contact us here. And be sure to attend “Seize the AI Moment,” our online event, on July 23rd where we have an all-star lineup to talk about what they’ve learned in serving this burgeoning market.
