Towards Responsive Data Centers Handling Massive Data

Abstract:
The common characteristic of today's data center applications is that they are data-intensive. Examples include Web search engines, social networking Web sites, in-cloud desktop software, and scientific applications that store, update, and process TBs to PBs of data. On the other hand, users demand responsiveness from these applications despite the scale of data that the applications need to handle.
Motivated by these observations, I believe it is imperative to design responsive data centers that handle massive data. This requires a comprehensive and principled examination of new problems (e.g., new data-processing methods such as MapReduce) as well as old problems (e.g., scheduling), treating responsiveness and data scale as the first-class metrics. In this talk, I will summarize my work and future directions along this vision.