Towards Task-Aware Data Center Networks

Abstract

Popular applications like Facebook and Google Search perform rich and complex "tasks" (e.g., generating a user's social newsfeed). From a network perspective, these tasks typically comprise multiple flows, which traverse different parts of the network at potentially different times. Existing network resource allocation schemes, however, treat all these flows in isolation - rather than as part of a task - which delays completion of tasks (i.e., user requests).

In this talk, I will show that task-aware network scheduling, which groups flows of a task and schedules them together, can reduce both the average as well as worst case task completion time for typical data center applications. I will also present the design, implementation, and evaluation of Baraat, a decentralized task-aware scheduling system that realizes these benefits for a wide range of real workloads and network settings. Finally, I will discuss how this work fits into my overall research agenda on improving various aspects of diverse networked systems.

Bio: Fahad Dogar is a postdoctoral researcher in the systems and networking group at Microsoft Research, Cambridge, UK. Earlier, he completed his PhD from Carnegie Mellon University and undergrad from LUMS, Pakistan. His research interests span the areas of mobile and wireless systems, Internet architecture and protocols, and cloud computing.