CSC 6991 - In-network Computing
Every Winter, Department of Computer Science, Wayne State University, Last Update, 2024
The next-generation hybrid cloud platform is integrated with Network Functions (NFs) across diverse network environments, ushering in a new computing paradigm known as in-network computing (INC). This introductory course on INC offers system and network engineers the opportunity to engage in large-scale system optimization. Participants will learn practical data plane programming skills, gain hands-on experience in hardware programming using software techniques, and explore real-world applications through case studies in data centers, distributed AI/ML systems, security, and more.
Why In-network Computing?
Background. The next-generation hybrid cloud platform has been chained with Network Functions (NFs) across various network settings. This technological progress has ushered in a new computing paradigm called in-network computing (INC).
Microsoft optimizes and scales next-generation 5G networks with INC Network engineers can implement various Network Functions on programmable switches or network interface cards (NICs), enabling the handling of large volumes of traffic (e.g., > 10 Tbps) in a cost-efficient manner (e.g., one programmable switch versus tens of servers) without needing to dedicate CPU cores specifically to network functions.
Nvidia achieved end-to-end AI system optimization with INC (Oct. 5. 2024)
https://developer.nvidia.com/blog/advancing-performance-with-nvidia-sharp-in-network-computing/
Course Overview
This is an introductory course to INC.
New Trends and Opportunities: Tech companies, including Google, Amazon, Microsoft, NVIDIA, and others, are significantly improving the performance of large-scale AI systems through network programmability. This rapidly growing field is a key focus of current research, also increasingly vital in the job market.
Practical Skills: Learn a straightforward yet powerful P4 language (C++ like) that enables engineers to customize hardware functions of commercial server’s Network Interface Cards (NICs) and switches to meet specific needs.
Hands-On Experience: The course provides easy access to the P4 programming, featuring progressive exercises and a virtual machine pre-equipped with all necessary software diving into data plane programming with the tools at your fingertips.
Case Studies: The course explores various real-world applications of INC. Gain the skills to effectively deploy the P4 program in addressing intricate challenges such as data center load balancing, network monitoring, and more. Deepen your understanding of P4’s potential and limitations through practical examples and critical analysis.
Topics
- Software-Defined Networking (SDN)
- Network Function Virtualization (NFV)
- Concept of In-network Computing (INC)
- Network programmability with P4
- Approximation algorithm for P4
- INC for security (botnet, DDoS, intrusions, etc.)
- INC for data center (load balancing, caching, etc.)
- INC for distributed systems (IoT, smart grid, etc.)
- INC for machine learning and AI (DT, RF, RNN, etc.)
Recent Information
- Term: Winter 2025
- Time: TT 11:30 AM-12:45 PM
- Room: 0301, STEM Innovation Learning (online attendance upon approval)
- Prerequisites: experience with Linux, c/c++ (or java, python)
- Registration due on Dec 15, 2024