It’s Chloe here, your trusty engineer-turned-biz-dev-manager. I’ve been working with the developer and partner ecosystem team in Cisco DevNet, and I’m especially excited to have the opportunity to reach the 2019 GPU Technology Conference developer audience with this post. I want to give you the low-down about the latest and greatest things that Cisco and the Cisco DevNet team are doing to empower developers such as you, specifically for analytics and data computation with Cisco Unified Computing System (UCS) servers.
First, what is Cisco DevNet?
Cisco DevNet is Cisco’s developer program. It gives you a centralized hub for all things developer at Cisco. For app developers and infrastructure programmers, from ‘hello world’ to automating the entire data center, or somewhere in between, DevNet offers you various resources to help you on your developer journey. There are learning labs, video courses, hands on environments called sandboxes, a place to post and share your code through what we call Code Exchange, and a place for our partner ecosystem to share solutions that drive business outcomes using Cisco technologies with Cisco customers in Ecosystem Exchange.
Now fill me in on Cisco UCS technology, please?
Since you said please – let me tell you about this amazing latest piece of technology released from Cisco, the UCS C480 ML. This bad boy is equipped with eight NVIDIAⓇ Tesla v100 GPUs. Can you give me a “cha-ching!” Additionally, it offers zero touch deployment and can be fully managed through APIs. You know about APIs…right? Means this server is programmable, but not just programmable, 100% automatable! The UCS Unified API provides full coverage of all the C480 ML configuration requirements. Everything from BIOS settings, to Virtual Drive configuration, from Boot policies to Virtual Media mappings. When your AI/ML workloads change, the UCS Unified API allows the C480 ML to adapt to the new requirements at the click of a button, the sound of your voice, the message on the bus, the CI in your CI/CD pipeline… I think you get the idea.
This powerful beast allows for Machine Learning and Deep Learning data computation to happen on-premises or through a hybrid cloud configuration. Using NVIDIA’s NGC containers, it has support for all major ML frameworks; Tensorflow, PyTorch, Caffe2, Keras, to name a few. Bonus – the C480 ML is also RAPIDS ready, as part of the NGC-Ready Validation. From data collection and analysis near the edge, to data preparation and training in the data center, to the real-time inference at the heart of AI, customers are covered.
Learn more about the UCS C480 ML.
Sounds great, now how do I get my hands on the UCS C480 ML?
Great question. As a Cisco customer you can reach out to your account team and they can hook you up. Want to try before you buy? – DevNet has got your back. DevNet recently stood up a DevNet Sandbox where you can access the UCS C480 ML via an online, reservation-based system (fo’ free!!). We’ve also put together a great learning lab to get you up and running. Here are instructions on how to reserve the sandbox.
Here are a few details of the UCS C480 ML Sandbox to get you excited. It contains an Ubuntu 16.04 VM with:
- 8x NVIDIA V100 Tensor Core 32GB GPUs in PCI Direct I/O Passthrough
- NVIDIA Drivers 410.104 with CUDA 10.0
- nvidia-docker
Make sure to the check out the DevNet UCS C480 ML with NVIDIA GPU Cloud Sandbox for the latest details.
The only catch – you have to become a member of Cisco DevNet… Okay, this isn’t a catch at all. It’s awesome, it’s free, but you don’t have to take my word for it. Sign-up and see for yourself.
Are you thinking, “What else should I check out?”
Since you’re interested, I have a few more resources that I’d recommend you take a peak at. There is a more technical blog about building retail experiences using machine learning which is the demo that we are showing at the 2019 NVIDIA GPU Tech Conference. Additionally, head over to the new Cisco Computing Solutions for AI & ML page for resources and information on the power of the possible. If you’re interested in the UCS programmability, check out our UCS Dev Center, as well as some Cisco UCS repos on Code Exchange.
And a couple other Cisco blogs about AI and ML:
- Overcoming Unpredictable AI Data Pipelines
- Performance Scale, and Flexibility for Accelerating AI/ML
- The AI Mandate: To Boldly Go Where No Data Center Has Gone Before
P.S. Cisco DevNet is hosting its third annual developer conference DevNet Create at the Computer History Museum in Mountain View, CA April 24-25, 2019. Use the promo code “NVIDIA” for 50% off the conference pass. I hope to see you there!
Get your free DevNet account for free access to resources, learning labs, and sandboxes for network automation and application development.
Chloe Kauffman on Twitter Follow @hchloemkauffman
What a well written article.