Advances in sensing technologies are yielding vast quantities of data that must either be processed on the fly or archived for later consumption, or both. Unfortunately, new applications, eager to harvest such data, are starkly limited by current techniques that tend to be hard to scale up, difficult to adapt, and largely batch-oriented. This paper discusses the research on key new cloud-computing technologies that will deliver higher performance for stored data at scale, lower latency for streaming data and a lower-cost network infrastructure.