Parallelization Strategies in Neural Networks

From a network engineer’s perspective, it is not mandatory to understand the full functionality of every application running in a datacenter. However, understanding the communication patterns of the most critical applications—such as their packet and flow sizes, entropy, transport frequency, and link utilization—is essential. Additionally, knowing the required transport services, including reliability, in-order packet delivery, and lossless transmission, is important.

In AI fabrics, a neural network, including both its training and inference phases, can be considered an application. For this reason, this section first briefly explains the basic operation of the simplest neural network: the Feed Forward Neural Network (FNN). It then discusses the operation of a single neuron. Although a deep understanding of the application itself is not required, this section equips the reader with knowledge of what pieces of information are exchanged between GPUs during each phase and why these data exchanges are important.


Feedforward Neural Network: Forward Pass


Figure 1-7 illustrates a simple four-layer Feed Forward Neural Network (FNN) distributed across four GPUs. The two leftmost GPUs reside in Node-1, and the other two GPUs reside in Node-2. The training data is fed into the first layer. In real neural networks, this first layer is the input Continue reading

AI Cluster Networking

Introduction

The Ultra Ethernet Specification v1.0 (UES), created by the Ultra Ethernet Consortium (UEC), defines end-to-end communication practices for Remote Direct Memory Access (RDMA) services in AI and HPC workloads over Ethernet network infrastructure. UES not only specifies a new RDMA-optimized transport layer protocol, Ultra Ethernet Transport (UET), but also defines how the full application stack—from Software through Transport, Network, Link, and Physical—can be adjusted to provide improved RDMA services while continuing to leverage well-established standards. UES includes, but is not limited to, a software API, mechanisms for low-latency and lossless packet delivery, and an end-to-end secure software communication path. 

Before diving into the details of Ultra Ethernet, let’s briefly look at what we are dealing with when we talk about an AI cluster. From this point onward, we focus on Ultra Ethernet from the AI cluster perspective. This chapter first introduces the AI cluster networking. Then, it briefly explains how a neural network operates during the training process, including an short introduction to the backpropagation algorithm and its forward and backward pass functionality.

Note: This book doesn’t include any complex mathematical algorithms related backpropagation algorithm, or detailed explanation of different neural networks. I have written a book Continue reading

The Rise of CLI-Based AI Coding Agents: Claude code vs Gemini CLI

Introduction I have been a Cursor user for vibe coding for 3 months. I was very skeptical about using Claude Code and Gemini CLI at first, since I wasn’t comfortable with the idea of using a terminal as an AI agent. But in the last 1–2 months, I’ve been trying them both — and it … Continue reading The Rise of CLI-Based AI Coding Agents: Claude code vs Gemini CLI →

How Agentic AI Is Redefining Campus and Branch Network Needs

The workplace is being redefined. AI workloads, an explosion of connected devices, and changing working patterns are forcing organizations to rethink their campus and branch network designs to support business goals and deliver great digital experiences to customers and employees. Over the last decade, IT teams have had to manage significant change with the adoption of cloud computing, widespread use of mobile devices, and SaaS applications becoming critical to core business operations. Now, the transformation that is AI presents an opportunity to gain a core competitive advantage and a productivity multiplier for those organizations that successfully embrace it. When it comes to the rise of Small Language Models (SLMs) and agentic AI, sophisticated AI capabilities are moving closer to where business happens — at the branch office and on campus. This shift to “edge AI” promises exciting possibilities but also brings significant implications for network infrastructure that network architects and decision-makers must address now. Understanding Local Small Language Models (SLMs) at the Edge Local SLMs are designed to be compact and efficient enough to run on local servers or even dedicated edge devices. For tasks like answering simple queries or summarizing documents using local data, these models perform inference right Continue reading

Congestion Control at IETF 123

As usual, IETF 123 was a busy week for DNS folk. I'll cover the material presented at the DELEG and DNSOP working groups. There is more to the DNS at IETF meetings than just these two working groups, and I'll skip over Adaptive DNS Discovery (ADD), Extensions for Scalable DNS Service Discovery (DNSSD), and DANE Authentication for Network Clients Everywhere (DANCE) in the interests of trying to keep this report (relatively) brief!

Congestion Control at IETF 123

The Internet Engineering Task Force (IETF) meets three times a year to develop Internet Standards and related best practices. At its July 2025 meeting in Madrid, several sessions explored the evolving role of congestion control in transport protocols and sparked the observations in this post.

Congestion Control at IETF 123

The Internet Engineering Task Force (IETF) meets three times a year to develop Internet Standards and related best practices. At its July 2025 meeting in Madrid, several sessions explored the evolving role of congestion control in transport protocols and sparked the observations in this post.

Calico at CalicoCon, KubeCon + CloudNativeCon North America 2025!

The Calico team was thrilled to participate in KubeCon + CloudNativeCon North America 2025, where we’ll be showcasing the latest advancements in Kubernetes networking, network security, and observability. We’re excited to connect with the vibrant cloud-native community, share insights, and demonstrate how Calico Open Source continues to empower organizations worldwide.

Quick Links

Nov 10: CalicoCon North America 2025

CalicoCon North America 2025, your go-to event for the latest in Kubernetes networking, network security, and observability.

Hosted by the Calico team, this hybrid event is your chance to hear directly from Calico engineers and leadership, get hands-on with new features, and take an in-depth look at the state of Project Calico.

We’ll dive into Calico 3.30, Calico eBPF, and Calico Whisker: open source observability for Kubernetes.

To view the full agenda & register, see below or click here:

CalicoCon 2025 Logo

Event Details

Date: November 10, 2025
Time: 1:00pm to 5:00pm EST
Location: Virtual | The Westin Peachtree Plaza Atlanta

Nov 10: Happy Hour with Calico

After a day of deep dives and technical Continue reading

MadeYouReset: An HTTP/2 vulnerability thwarted by Rapid Reset mitigations

On August 13, security researchers at Tel Aviv University disclosed a new HTTP/2 denial-of-service (DoS) vulnerability that they are calling MadeYouReset (CVE-2025-8671). This vulnerability exists in a limited number of unpatched HTTP/2 server implementations that do not sufficiently enforce restrictions on the number of times a client may send malformed frames. If you’re using Cloudflare for HTTP DDoS mitigation, you’re already protected from MadeYouReset.

Cloudflare was informed of this vulnerability in May through a coordinated disclosure process, and we were able to confirm that our systems were not susceptible, due in large part to the mitigations we put in place during Rapid Reset (CVE-2023-44487). MadeYouReset and Rapid Reset are two conceptually similar HTTP/2 protocol attacks that exploit a fundamental feature within the HTTP/2 specification: stream resets. In the HTTP/2 protocol, a "stream" represents an independent series of HTTP request/response pairs exchanged between the client and server within an HTTP/2 connection. The stream reset feature is intended to allow a client to initiate an HTTP request and subsequently cancel it before the server has delivered its response.

The vulnerability exploited by both MadeYouReset and Rapid Reset lies in the potential for malicious actors to abuse this Continue reading

Transform ISP Choice from Anecdote to Evidence

Have you ever had to defend your choice of internet service provider? All you can say is: “Everyone says they’re reliable”, “My buddy recommended them.”, “They are the incumbent player”. But when pressed about frequent connectivity issues, then what? Sound familiar? This plays out in businesses across the world every day. We make one of […]

The post Transform ISP Choice from Anecdote to Evidence first appeared on Rick Mur.

N4N036: OSPF Area Types

Ethan and Holly bring you the last installment of the OSPF series discussing OSPF area types. They discuss why OSPF areas exist, do a quick recap of what OSPF areas actually are, and then introduce the different types of OSPF areas.  Lastly, see if you can answer Ethan’s rapid-fire OSPF questions. Episode Transcript: This episode... Read more »

Worth Reading 081325


The current state of digital identity is a mess. Your personal information is scattered across hundreds of locations: social media companies, IoT companies, government agencies, websites you have accounts on, and data brokers you’ve never heard of.


It turns out that, if you have your domain hosted by a big provider (we happen to use GoDaddy), it’s easy to turn on DNSSEC. But I think it says a lot that it took us this long (and the stimulus of working on a new security book) to get us to turn on DNSSEC


As we left the laboratory, I thought about how we in the computing field build a tremendous number of things that really cannot be called beautiful and then are commonly tossed aside without a thought.


The accelerated migration to advanced services will be accompanied by unprecedented complexity, and security and reliability concerns that must be addressed by the network-engineering and formal-methods communities.


Quantum scientists have long treated quantum entanglement as precious cargo, forging fresh links for every secure message or computation. A new theoretical study proposes a thriftier route, letting an existing pair pass portions of its entanglement down an extended chain.

1 … 87 88 89 90 91 … 3,894