Systematic Analysis of Networked Systems
Event details
| Date | 15.01.2019 |
| Hour | 10:00 › 12:00 |
| Speaker | Luis Pedrosa |
| Location | |
| Category | Conferences - Seminars |
Abstract
Designing and maintaining networked systems poses many challenges to both developers and operators alike. Unexpected behavior can often result from subtle interactions between the networked entities and better tools are needed to help understand the underlying complexity.
The stakes are also ever increasingly higher, as more and more of our lives depend on the reliable, correct, and performant operation of these networked systems.
In this talk I will show how we can address these challenges by adapting techniques borrowed from the programming languages community. I will focus on two analysis frameworks I have developed: SPA and CASTAN. SPA is a framework that facilitates the systematic exploration of conversations in distributed systems, allowing users to check safety properties while considering all possible inputs and network dynamics.
CASTAN analyzes performance in network functions and automatically generates adversarial workloads that significantly slow down packet processing, allowing developers to debug such issues and operators to provision resources defensively. I will also talk about ongoing work that reasons more generally about network function performance with performance contracts.
Biography
Post-Doctoral Researcher, Network Architecture Lab, EPFL
Luis Pedrosa is a Post-Doctoral Researcher at the Swiss Federal Institute of Technology in Lausanne (EPFL). Before joining EPFL, he received his PhD in Computer Science at the University of Southern California in 2016.
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 analyzing the performance footprint of network functions to automatically derive poorly performing adversarial workloads and verifying their correctness and reliability. 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.
Designing and maintaining networked systems poses many challenges to both developers and operators alike. Unexpected behavior can often result from subtle interactions between the networked entities and better tools are needed to help understand the underlying complexity.
The stakes are also ever increasingly higher, as more and more of our lives depend on the reliable, correct, and performant operation of these networked systems.
In this talk I will show how we can address these challenges by adapting techniques borrowed from the programming languages community. I will focus on two analysis frameworks I have developed: SPA and CASTAN. SPA is a framework that facilitates the systematic exploration of conversations in distributed systems, allowing users to check safety properties while considering all possible inputs and network dynamics.
CASTAN analyzes performance in network functions and automatically generates adversarial workloads that significantly slow down packet processing, allowing developers to debug such issues and operators to provision resources defensively. I will also talk about ongoing work that reasons more generally about network function performance with performance contracts.
Biography
Post-Doctoral Researcher, Network Architecture Lab, EPFL
Luis Pedrosa is a Post-Doctoral Researcher at the Swiss Federal Institute of Technology in Lausanne (EPFL). Before joining EPFL, he received his PhD in Computer Science at the University of Southern California in 2016.
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 analyzing the performance footprint of network functions to automatically derive poorly performing adversarial workloads and verifying their correctness and reliability. 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.
Practical information
- General public
- Free
Contact
- EDIC - [email protected]