Jason Stowe, Cycle Computing | AWS Summit 2014
summary
Cycle Computing demonstrates utility supercomputing on AWS with a 156,000-core run that finishes weeks of work in hours — 18 hours for a material‑science job that would cost ~$68M on owned hardware but runs for $33,000 — and routinely builds clusters of 10k–50k cores and pilots that turn 20‑day runs into one‑day results with 25,000 cores. They ship a Data man scheduler for timed, protocol‑agnostic data movement (encrypting genomes into Glacier) and package Schrodinger Material Science as a one‑button large‑cluster offering.
AI-written summary · about
