Provide intuition on expected scale

[willscott]

It would be good for intuition of where this can be used to provide some expectations of how the service scales.
I think the specific things I'd be interested in are:

• does the database need to fit (in memory) on a single machine for practicality?
• what level of IO and computation are needed as usage and database size scale?
• what amount of tuning is needed for different read/write workloads?