Online / 6 & 7 February 2021


Datacenter class containers for the masses

What if we told you that there is a way to build a powerful kubernetes cluster that can perform data-center-class workloads in a form factor that you can run at your home, under your lab bench, or at a conference. What if we further told you that this hardware comes from a sustainable supply chain that will lower the carbon footprint of computing. This talk will detail an open hardware design, built for a "home" form factor, with a chassis that can hold from 4 to 12 nodes; that contains almost 100 compute cores; 2TB of memory; and over 30 TB of storage. Join us for an overview through photos, video demonstration, and chats with the designers.

Perhaps you work with scalable computing technology every day in far-off data centers or hosted in public clouds. What if you could have similar computing capability in your home, in your home lab, or in the premises of your small organization or business? What if you could run 1000s of containers in under 2000 watts? There are many reasons, from latency limitations, to cost constraints, to concerns over data sovereignty that might require the deployment of on-premise computing. We will overview and demonstrate a new computing form factor that brings a data center class cluster using hyperscale technology that you can plug into your home's power outlet. We also integrate these clusters using materials from the circular economy, building on a sustainable supply chain that decomposes servers from top tier data centers and builds new products based on open designs. The power of open hardware and open software creates the possibilities of executing many compute intensive workloads in diverse locations without the need of a data center. This talk will demonstrate on open hardware design for a powerful deskside chassis that can hold from 4 to 12 nodes of scalable computing running from a single 110V/220V outlet and hosting a range of elastic k8s container workloads. Our short talk will include video demonstration, photos, and Q&A with the system designers.


Photo of Sri Ramkrishna Sri Ramkrishna