CS Colloquium - Anuj Kalia
Refreshments available at 10:15
Host: Richard Yang
Title: Software-optimized Systems in the Era of Hardware Specialization
Abstract:
In the post-Moore era, using specialized hardware tuned to specific applications is a promising direction to get higher performance. To create a cohesive specialization roadmap for the future, we must ask: how far can we go by optimizing software for existing hardware, and when must we turn to the more expensive option of deploying specialized hardware? I argue that for many important systems problems for which specialized hardware—intelligent NICs, FPGAs, programmable switches, and GPUs—has been proposed, software-optimized systems can provide competitive performance.
In this talk, I will focus on this optimization-specialization tradeoff in the context of distributed systems for modern datacenter networks. By using new optimizations based on an improved understanding of datacenter hardware, my work invalidates the commonly-held belief that software-based networking cannot match datacenter network speeds. This allows building fast distributed systems that run entirely on existing CPUs. I show that such designs have a fundamental latency advantage over distributed systems that use specialized hardware, whose limited flexibility often increases network round trips. I will describe in detail the design of eRPC, the first networking library to provide near-network performance in commodity datacenters. eRPC aligns with the end-to-end principle, and answers long-standing networking questions about reliability and congestion control. Finally, I will discuss how by better leveraging existing hardware, and carefully choosing hardware specializations, we can create future systems that match the performance of ever-faster networks.
Bio:
Anuj Kalia is a PhD candidate in the Computer Science Department at Carnegie Mellon University, advised by David Andersen and Michael Kaminsky. His research interests are in high-performance computer systems, with a focus on distributed systems for modern datacenter networks. He is a recipient of a Facebook PhD fellowship and a USENIX ATC Best Student Paper award.