Design a restricted runtime model so small short-lived services can be deployed with less latency.
I would like to be able to rely on Azure to run very short but highly parallel jobs that might, say, consume 1000 CPUs for 30 seconds each. Startup costs make this prohibitive.
Consider how useful this would be for the problem of doing a custom semantic web search, or searching through a large configuration space of some simply defined problem for the configuration which yields the highest utility.
8
votes
anonymous
shared this idea
