The Scale of the Cloud

Each task has its own docker container. And then we push those out into the cloud, and they get allocated out on to these machines. They run their task, data comes in from s three, goes back out to s three. Nothing is left of the hard drives, and then they disappear thee. The're these little ephemeral compute instances. We do occasionally have a certain type of molecular modelling job that we can scale very wide. I think our largest jobs to day have been about ten thousand s ps wide and running for a few days. So maybe like four or five days.