Fundamentals of Real-Time & High-Performance Computing on Linux

General course information

Lecturer: Prof. Tommaso Cucinotta.
Duration and format: 20 hours of lab lessons (2 CFU).
Schedule: TBA
Location: TBA
Exam: coding project and oral examination.

Description

The course aims at providing basic knowledge on mechanisms available on Linux to control and optimize the performance of real-time and high-performance applications.

Program at a glance

The following topics will be covered:

optimization of C/C++ code, compiler options, impact of tracing, logging and assertions, profiling and optimization of the branch predictor;
memory allocation, swap control and control of the interferences due to memory handling at the OS/kernel level;
DVFS and deep idle states of modern CPUs, their impact on performance and latency, and how to control them;
core and IRQ pinning;
NUMA architectures;
scheduling for real-time applications, how to tune the POSIX priority-based scheduler, and the SCHED_DEADLINE one available in mainline;
high-performance inter-process communications and their optimization through busy waits, spin-locks and atomic variables;
tracing sources of "OS noise" using perf and oslat;
kernel bypass techniques for high-performance networking exploiting such frameworks as DPDK.

About the lecturer

Prof. Tommaso Cucinotta has a MSc in Computer Engineering from University of Pisa and a PhD from Scuola Superiore Sant'Anna. He spent more than 10 years at the Real-Time Systems Laboratory (ReTiS) of Scuola Superiore Sant'Anna carrying out research in security and smart-card based authentication, adaptive deadline-based scheduling in the Linux kernel for embedded, soft real-time and multimedia applications, temporal isolation in virtualized cloud services and novel OS designs for massively parallel and distributed systems. He has been MTS at Bell Labs in Dublin, carrying out industrial research on security and confidentiality, and real-time performance of cloud systems, with a focus on Telco applications. He has also been a Software Development Engineer in AWS DataBase Services in Dublin, Ireland, working on scalability and performance enhancements to the AWS DynamoDB NoSQL real-time data store. Since 2016, he is back at the ReTiS of Scuola Superiore Sant'Anna as associate professor. He is a member of the PhD board of the Data Science PhD program jointly offered by Scuola Sant'Anna, University of Pisa, Scuola Normale Superiore, IMT Lucca and CNR since its first edition in the a.y. 2017/18. He is also a member of the PhD board of the National PhD in AI - AI for Society - program jointly offered by Scuola Sant'Anna, University of Pisa and other academic institutions since its first edition in the a.y. 2021/22. Since year 2019, he is head of the real-time and embedded systems research area at the RETIS.

Last updated on
04 October 2024