All Posts

AI infrastructure, DC fabric design, and the CCDE journey. Published every two weeks.

AI Infrastructure
Rail-Optimized or Full CLOS: How Should You Build a 2,000-GPU AI Fabric?
The topology decision you make early in a GPU cluster design will follow you for years. Rail-optimized and full CLOS are not opposites — the real question is where you want the complexity to live.
2026-09-03 · 6 min read
→
AI Infrastructure
AI Is Forcing Ethernet To Become Something It Was Never Designed To Be
For decades Ethernet accepted packet loss as normal. Now the largest AI environments in the world are spending significant engineering effort trying to avoid it. That shift is more interesting than it sounds.
2026-06-03 · 5 min read
→
AI Infrastructure
A Dropped Packet in an AI Training Cluster Is Not the Same as a Dropped Packet Anywhere Else
In traditional networks TCP recovers. In AI training clusters the entire job waits. Understanding why changes how you think about fabric design completely.
2026-05-27 · 5 min read
→
CCDE Journey
CCIE vs CCDE — The Justification Is the Answer
The CCDE is not a harder CCIE. It's a completely different question being asked. And the answer is never the answer — the justification is.
2026-05-19 · 6 min read
→
CCDE Journey
The CCDE Got Electives. Almost Nobody Is Talking About It.
Cisco updated the CCDE over a year ago and added electives. One of them is AI Infrastructure. Almost nobody in traditional networking knows it exists.
2026-05-08 · 4 min read
→