I had the privilege to get outside of my normal daily opportunities and learn about a new industry that I knew little about, AI Infrastructure. The folks at Tech Field Day invited me to AI Infrastructure Field Day where I got to meet some companies and learn about a whole new industry. My biggest take away from Hammerspace’s presentation was:
Hammerspace’s integration into the linux kernel is a game changer compared to the others. I’m a big fan of following standards. I’m excited to learn more about that piece.
So how does following standards help the AI Infrastructure industry?
Through eliminating data silos across your cloud/hybrid/on-premise data centers and edges.
These companies are not your typical AI companies that get all the hype. Everyone is adding AI capabilities to their products, and how many times have I heard about the fatigue of AI across industries. But how is AI being added on the backend? That is where AI Infrastructure Field Day enters the arena.
When participating in AI Infrastructure Field Day 3, there was a regular terminology that came up revolving around the AI models. The one stand out was how to get your data into the AI models being processed by Nvidia GPUs? Or how are you building GPUs as a service.
This is where Hammerspace really shined, because they are following a standards approach and getting things built into the Linux Kernel instead of relying on another client to be installed on your servers. That is the game changer and why many people are selecting Hammerspace for the AI Infrastructure needs.
I’ve attended Edge Field Day events in the past where we focused on the data that is being collected and processed at the Edge of our networks. Many of the AIIFD companies were focused on that Edge as well.
Enterprises can deploy Hammerspace and get all their data into a central location. AI is data heavy and loves everything it can get and process. So bringing it together into a central location is a key step in the journey. Once you have a central location, enterprises can plan migrations with the future in mind. This is helpful for migration between infrastructures or new sites. Once again we come back to the standards view with Hammerspace excelling in this point.
Data Sprawl is the bane of any large enterprise. Hammerspace brings a way to manage and handle all that data.
Hammerspace is building their client into the Kernel so you don’t have to install and manage any additional software. This is being pushed through NFSv4.2. Something already being build into the Linux Kernel and has wide support across the industry.
They are also pushing for a new modular serve model using NVMe storage. It’s an interesting view to take and hopefully they get some uptake of this new form factor.
Instead of relying on proprietary ways of handling the data and getting it to a central location for the AI Training and Inference processes, Hammerspace is building something that can be used by everyone. That is the kind of technologies that benefit society.