At the SC25 event, DriveNets’ Dudy Cohen (VP Product Marketing) and Andy Holland (Sales Director for AI Infrastructure) explore how Ethernet is expanding its reach beyond traditional scale-out and scale-across applications to challenge InfiniBand in the demanding scale-up domain.

Key highlights

• DriveNets powers the world’s largest AI and Telco networks
• DriveNets brings an AI Networking solution with peak performance out-of-the-box
• DriveNets fabric scheduled Ethernet is solving industry-wide pain points in AI Networking

Watch the interview 

Want to learn how DriveNets is reshaping AI networking infrastructure?
Explore DriveNets AI Networking Solution

𝗪𝗵𝗲𝗿𝗲 can you 𝗳𝗶𝗻𝗱 𝗗𝗿𝗶𝘃𝗲𝗡𝗲𝘁𝘀
LinkedIn
X

What is the new service provider network? This is the network modernization process all service providers need to go through. Service Providers cannot continue to build networks the same way they have for over 20 years, and still maintain a valid business.

DCI is going through something and this something is called AI. The three changes that AI made to DCI are capacity, performance, and one environment and layer three capabilities.

What’s the importance of latency in AI networks. AI networks introduce new challenges that need different treatments of latency.

DES, DSF, DDC – Comparing Scheduled Fabric
Arista recently launched their DES solution, the Distributed Etherlink Switch, which is essentially an end to end VOQ system – a large scale chassis. The approach is basically saying we should put as much logic as we can on the switch side. The switch will handle all the congestion control, all the reordering of packets, the load balancing of packets, which are all necessary for AI networking. So why choose DriveNets DDC over those solutions?

Are autonomous networks coming soon?
We’re taking baby steps, starting with islands of automation and we’re going to automate small, mundane tasks by helping operators develop CI/CD pipelines that allow them to take innovation from their lab to a network digital twin and eventually into production in a much faster way.

We have an issue with resolving the AI fabric, or an AI networking problem with large clusters of GPUs usually used for training. This episode looks at the issue with resolving the AI fabric and explores how an Ethernet based solution can resolve this – by building a chassis which is distributed (a disaggregated, distributed chassis). This approach has no packet loss, is lossless and is a fully scheduled fabric, but without the scale limitation of a chassis.

What are the new automation sets coming to DriveNets Network Orchestrator?

Autoboot profiles helps you bootstrap new white boxes from the factory. Smart Rollout is a brand new feature and it allows you to push software into the entire network after configuring just a few parameters. The Single Source of Truth integration seamlessly integrates the inventory that we have in DNOR into NetBox, or or other inventory systems.

Failure recovery is a very big issue when it comes to AI clusters because there are always failures and when the failure come, it’s a big thing because you need to stop the calculation, go back to the last checkpoint. You lose a lot of time and money and resources that are spent idle and and wasted time. And the networking part is crucial in order to create a fail.

Congestion demands attention, otherwise it can result in higher latency and packet loss. The main dilemma around congestion is around avoidance or mitigation. We’ll look at how scheduled fabrics make sure that AI infrastructure is lossless and predictable rather than bringing additional technologies to mitigate the congestion, but do not avoid it