This is not "AI for network engineers" but rather "Network engineering for AI datacenters". I was expecting to read that a small neural network could be used to direct traffic.
Far more practical than I expected. I particularly enjoyed the detailed diagrams.
"Though BGP supports the traditional Flow-based Layer 3 Equal Cost Multi-Pathing (ECMP) traffic load balancing method, it is not the best fit for a RoCEv2-based AI backend network. This is because GPU-to-GPU communication creates massive elephant flows, which RDMA-capable NICs transmit at line rate. These flows can easily cause congestion in the backend network."
For a latest reference on AI and machine learning for network engineer please check this book by Javier Antich [1].
Please also check the review here [2]. For what it's worth, the book is listed in the "10 Books Every Network Engineer Should Read" [3].
[1] Machine Learning for Network and Cloud Engineers: Get ready for the next Era of Network Automation:
https://www.goodreads.com/book/show/101180344-machine-learni...
[2] MUST READ: Machine Learning for Network and Cloud Engineers:
https://blog.ipspace.net/2023/02/machine-learning-network-cl...
[3] 10 Books Every Network Engineer Should Read:
https://networkphil.com/2024/05/21/10-books-every-network-en...