Posts on Sebastian Scheinkman - Red Hat Openshift, Networking, Kubernetes and Cloud Native

Bridging Worlds

Sun, 12 Oct 2025 00:00:00 +0000

Bridging Worlds: High-Performance Telco Networking in Kubernetes with SR-IOV, DRA, and NRI

The Telco Imperative: Why Bare-Metal Speed Matters in a Virtualized World

The telecommunications industry is undergoing a profound transformation, driven by the shift from proprietary, hardware-centric appliances to flexible, software-based solutions running on standard IT servers. This paradigm, known as Network Functions Virtualization (NFV), promises unprecedented agility, scalability, and cost efficiency. However, this transition introduces a formidable challenge: performance. For critical network functions, particularly those in the data plane of a 5G network like the User Plane Function (UPF), performance is not merely a feature—it is an absolute prerequisite. These functions must process millions of packets per second with minimal latency, a task for which traditional software-based virtual networking is often ill-equipped. This is where Single Root I/O Virtualization (SR-IOV) becomes an indispensable technology.

Beyond eth0

Sat, 04 Oct 2025 00:00:00 +0000

Beyond eth0: A Deep Dive into Kubernetes Multi-Network Pods with Multus CNI

The Kubernetes Network Model: A Foundation of Simplicity and its Limitations

At the heart of Kubernetes’ design is a networking model that is both powerful in its simplicity and opinionated in its structure. This model was created to solve the complex networking challenges inherent in running distributed systems across a cluster of machines. To truly grasp the need for advanced solutions like Multus, one must first understand the foundational principles of Kubernetes networking and the specific limitations that drive the demand for multi-homed Pods.

Unlocking AI at Scale

Sun, 24 Aug 2025 00:00:00 +0000

Unlocking AI at Scale: A Deep Dive into RDMA, InfiniBand, and RoCE with NVIDIA Mellanox

The exponential growth in the scale and complexity of artificial intelligence models, particularly Large Language Models (LLMs), has created an unprecedented communication bottleneck in distributed computing systems. As these models expand beyond the memory capacity of a single GPU or even a single server, they necessitate multi-node clusters where efficient inter-node communication is paramount. In this high-stakes environment, traditional networking stacks like TCP/IP, which have served as the backbone of the internet for decades, are no longer sufficient for the demands of modern AI workloadshe overhead associated with CPU-managed data transfers and protocol processing introduces latency that can cripple the performance of tightly coupled GPU clusters.

The Ultimate Guide to tmux: Supercharge Your Terminal Productivity

Mon, 07 Jul 2025 00:00:00 +0000

The Ultimate Guide to tmux: Supercharge Your Terminal Productivity

Introduction: Taming Terminal Chaos

For any professional who lives in the command line—be it a software developer, a DevOps engineer, or a system administrator—the terminal is both a sanctuary and a potential source of chaos. A typical workflow might involve a dozen terminal tabs: one running a local web server, another for the front-end build process, a third for a database console, and several more for various microservices. Add to this a few separate SSH windows connected to staging and production servers, where a critical, long-running data migration is underway. The digital workspace is a fragile tapestry of processes, and the constant fear lingers: a dropped Wi-Fi connection, an accidental window closure, or a system reboot could bring the entire house of cards crashing down, losing hours of work and context.

Unleashing Bare-Metal Performance in Kubernetes

Sun, 29 Jun 2025 00:00:00 +0000

Unleashing Bare-Metal Performance in Kubernetes: A Technical Deep Dive into the SR-IOV Network Operator

I. Introduction: Beyond Standard Kubernetes Networking

Standard Kubernetes networking, a cornerstone of its flexibility and portability, is primarily designed for general-purpose connectivity. It typically relies on software-based overlay networks, such as those using VXLAN, and kernel-level packet processing. While this model excels at providing seamless pod-to-pod communication and service discovery across a cluster, it introduces inherent latency and CPU overhead. For a vast majority of applications, this performance profile is more than sufficient. However, for a growing class of high-performance workloads, this software-defined approach represents a significant performance ceiling.