Implementation of a Prediction and Shared Buffer Management Algorithm in Programmable Switches

• Grant number: 24/18365-0
• Support Opportunities: Scholarships in Brazil - Master
• Start date: December 01, 2024
• Status: Discontinued
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Principal Investigator: Fabio Luciano Verdi
• Grantee: Mateus de Nuzzi Bragatto
• Host Institution: Centro de Ciências em Gestão e Tecnologia (CCGT). Universidade Federal de São Carlos (UFSCAR). Campus de Sorocaba. Sorocaba , SP, Brazil
• Company: Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia Elétrica e de Computação (FEEC)
• Associated research grant: 21/00199-8 - SMART NEtworks and ServiceS for 2030 (SMARTNESS), AP.PCPE
• Associated scholarship(s): 25/09024-7 - Implementation of a Prediction and Shared Buffer Management Algorithm in Programmable Switches, BE.EP.MS

Abstract

The master's project titled "Implementation of a Prediction and Shared Buffer Management Algorithm in Programmable Switches" aims to investigate the integration of a shared buffer management algorithm in programmable switches, focusing on the Tofino switch architecture and the use of the P4 language (Programming Protocol-independent Packet Processors). The proposal involves implementing the Credence algorithm, which uses predictions based on machine learning techniques to optimize the management of shared buffers in these devices. The Credence algorithm, developed to be implemented in simulated environments with the possibility of development in real hardware, seeks to improve the efficiency of network packet handling by applying prediction techniques to simulate the execution of a well-known algorithm from the literature: the Longest Queue Drop (LQD), which enables intelligent decisions about the acceptance or discard of packets before they are queued. Among the approaches explored for its implementation, packet recirculation, the use of ghost threads, and the implementation of virtual queues (vQueues) stand out, aiming to simulate and optimize the shared buffer occupancy and throughput performance. The research is organized into stages that include a literature review, the implementation of the mentioned techniques, and the evaluation of results through experiments on Tofino hardware. Additionally, the project includes a research internship abroad in collaboration with the Internet Architecture and Management Research Group (INET) at Technische Universität Berlin (TU Berlin), under the supervision of Prof. Dr. Stefan Schmid, an expert in computer networks and cybersecurity.