A down-to-earth look at the cloud host OS

Abstract

Current cloud programming models have opened up new opportunities, but the platforms they run on are still rooted in the legacy of single machine-centric computing. This leads to inefficiency that both costs money and offends scientific sensibilities. In this position paper, we make a passionate and necessarily opinionated argument for a research agenda that challenges fundamental assumptions about operating systems and "cloud" application software. We present a set of ideas for possible directions, and hope to elicit fruitful discussion within the community.