Unleashing Hardware Potential through Better OS Abstractions
Speaker: Adam Belay, Stanford University
Location: Warren Weaver Hall 1302
Date: March 4, 2016, 11:30 a.m.
Host: Subhash Khot
Datacenter workloads have demanding performance requirements, including the need for high throughput and low tail latency while maintaining high server utilization. While modern hardware resources are compatible with these goals, overheads and inefficiencies in today's operating systems remain a critical bottleneck. Significant improvements are possible through better operating system abstractions, but software and hardware complexity makes it difficult to deploy kernel changes in practice. I will first discuss Dune, a system that bypasses complexity by making it possible to run operating systems inside ordinary Linux processes. As a result, developers can use Dune to safely and efficiently access the raw capabilities of hardware while also falling back on all of the functionality and convenience of a full Linux environment. I will then discuss IX and Shinjuku, two generations of new datacenter-focused operating systems that were enabled by Dune. IX provides a novel system call interface that greatly improves network throughput without sacrificing latency, and Shinjuku uses a centralized approach to intra-server load balancing in order to dramatically increase CPU utilization.
Adam Belay is a Ph.D. candidate in Computer Science at Stanford University, where he is a member of the Secure Computer Systems Group and the Multiscale Architecture and Systems Team. Previously, he worked on storage virtualization at VMware Inc. and was a major contributor to the Linux Kernel project. His research area is operating systems and networking with the goal of restructuring computer systems so that developers can more easily reach the full performance potential of hardware. Adam has received the Stanford Graduate Fellowship, the VMware Graduate Fellowship, and an OSDI Jay Lepreau Best Paper Award.
Refreshments will be offered starting 15 minutes prior to the scheduled start of the talk.