Using cloud can speed up your time to science. Why every scientist (and citizen) should care.

In IT (and especially in research IT) we have a pantheon of metrics that we use daily, weekly, and monthly to assess the performance of our large-scale compute systems. CPU utilization, memory bandwidth per core, Lustre filesystem throughput - all the way down to cents per core-hour and data center PUE (which boils down to a measure of the thermodynamic efficiency of heat pumps). But there’s a dog not barking: how do we focus on scientist productivity, rather than machine efficiency?