DATE: Thursday,September 22, 2022
TIME: 12:00 - 1:00 pm
RMCIC Building, 4th Floor Panther Hollow Conference Room (4105)
SPEAKER: Luis D. Pedrosa, CMU Visiting Faculty, Instituto Superior Técnico, University of Lisbon, Assistant Professor
TITLE: Automated Reasoning about Networked Systems
In this presentation I will outline a high level overview of my research in building tools that reason about the performance, correctness, and reliability of modern networked systems. I will then take a deeper dive into CASTAN, a tool to help debug performance issues in software network functions (NFs). Running NFs on commodity servers simplifies the deployment of network services and reduce network operation cost, but raises the issue of unpredictable performance. Knowing this, network operators need to consider the performance of the deployed NF not only for typical but also adversarial workloads. CASTAN helps solve this challenge: it takes as input the LLVM code of a network function and outputs packet sequences that trigger slow execution paths. Under the covers, it combines directed symbolic execution with a sophisticated cache model to look for execution paths that incur many CPU cycles and involve adversarial memory-access patterns. We used our tool on 11 network functions that implement a variety of data structures and discovered workloads that can in some cases triple latency and cut throughput by 19% relative to typical testing workloads.
Luis Pedrosa is an Assistant Professor in the Department of Computer Science and Engineering at Instituto Superior Técnico / University of Lisbon and a Researcher at INESC-ID. Before joining IST, he received his PhD in Computer Science at the University of Southern California in 2016 and was subsequently a Post-Doctoral Researcher at the Swiss Federal Institute of Technology in Lausanne (EPFL).
With work spanning the systems, networking, and programming languages communities, Luis' main research interests concern the novel applications of PL techniques towards the understanding and improvement of modern networked systems. His recent work involves synthesizing enhanced network functions for network accelerators and reasoning about their correctness, reliability, and performance. He has prior work in verifying and analyzing distributed systems, but has also worked in a wide variety of other fields, including large-scale cluster management, mobile and cloud computing, low-power embedded systems, design automation, and even reverse engineering automotive electronic control units.
VISITOR HOST: Justine Sherry
VISITOR COORDINATOR: Sara Golembiewski, (for any non-seminar help / meeting help if needed)
SDI SEMINAR QUESTIONS?
Karen Lindenfelser, 86716, or visit www.pdl.cmu.edu/SDI/