How Brae uses Plexus

We run GPUs in an underwater data center. Our main focus is validating our hardware in the field, so when we thought about the software we needed, we could build it in house or find a third party like Plexus. It was a pretty easy decision. At our stage we need to move fast and spending time building and maintaining software was not something we wanted to do.

Mariah (CTO) and I have built software companies before, so we could have developed our own telemetry stack. It would have taken an engineer about a month to get something working, and then someone would have to own it for the life of the company. Every new sensor or view would have meant more work maintaining and updating the system instead of working on the pod.

Choosing Plexus let us spin up all our storage, monitoring, and visualizations in two days, so we could move much faster on our hardware. On Plexus we see all our data together the pod’s motion, the cooling loop, and the power, health, and temperature of the servers.

We also wanted to work with a startup. When we need something, I text Ann and it gets built. When we signed our first customer, we needed on-call schedules and phone call alerts in case there was an issue with the pod. Less than three days later our on-call schedule was set up. With a bigger company we would not get the same kind of attention.

When we started with Plexus we were running prototype tests. Three months later we have a production rack running for a customer, monitored the same way.

Vishnu Indukuri, CEO, Brae Systems