Fifty-plus EC2 compute nodes were the starting point.
Peek was running in Amazon, but I was pricing the first physical footprint for the company. The initial plan was a half cabinet with two VMware hosts and a pair of firewalls, switches, and routers.
The form asked for an estimated start date. I wrote “Immediately.”
I wanted to know how the data center handled bandwidth, power, contract terms, and growth. Ten megabits per second could be enough to begin, but only if the service could expand without turning every change into a negotiation. I asked what it would take to move from half a cabinet to a full one and then to several cabinets.
The vendor asked how large I expected Peek to become.
“Sky is the limit on growth,” I wrote.
That was enthusiastic, but the infrastructure plan was not just optimism. Peek's devices depended on services that connected mobile networks, outside email providers, and the company's own application. The routing and security equipment had a different job from the elastic compute nodes in Amazon. I wanted a small physical base that could connect those pieces and grow if the consumer business worked.
My rough horizon was twelve months. We could start with a little equipment and move more of the operation off EC2 as the requirements became clear. I could imagine the half cabinet eventually becoming a 300-to-500-square-foot cage.
I had seen that progression at Beatport. Small decisions become the permanent architecture quickly. A cheap cabinet is only cheap if the power, network, remote access, and contract still work when the company adds the next rack.
The quote itself was not the decision. It was the first useful set of numbers.
I had given notice at Yodle the day before and had not officially started at Peek. The data center planning was already underway.