The ask from the business
Cisco IT needed to build an AI cluster to support model training and inferencing for multiple teams across the business.
Challenge
Cisco product teams needed a way to run AI workloads that would be used to develop and test new AI capabilities in Cisco products. But, existing data center facilities weren’t built to power modern AI.
- Critical business need with tight timelines: Product teams were under pressure to deliver innovations to customers and needed the environment in 3 months.
- Extensive technical requirements: Needed to ensure security, reliability, and performance within the AI infrastructure.
- Hurdles to overcome: There were several challenges to figure out along the way related to power limitations, structured cabling limitations, and the location of data.
- Designing for the future: Needed to design to scale with the business, including automated operations.
- Building from scratch with a full data center refresh would take 18-24 months—we needed it done much faster.
Solution
Cisco IT designed an AI infrastructure with Cisco compute, best-in-class GPUs from NVIDIA, and Cisco networking. They built a front-end ethernet network and a backend lossless ethernet network — ensuring a reliable, high performing, scalable network.
Hear from the team
Jon Woolwine, Technical Systems Engineering and Jag Kahlon, Principal IT Engineer, explain how this new infrastructure was launched in only 3 months and expanded to support dozens of use cases.
