Oluwatosin Oladayo Aramide

vJunos on containerlab is routine on Intel. On an AMD EPYC host the stock image refuses to start its nested VM, and the first test topology most people write hides a second failure that looks like AMD and is not. What breaks, what current vrnetlab already fixes, all reproduced on an EPYC 9575F.

Put the vJunos 26.2 images on an AMD EPYC box running EVE-NG: the switch and router power themselves off, patch that and they hang at mount root, and Evolved on the default AMD template never starts QEMU. Four small edits in two places, each reproduced on the EPYC, every command ready to copy.

Part 4 of the Agentic NOC series. A cited timeline: MCP won the vertical axis in a year, A2A won the horizontal axis with a hundred-company tent, and ACP converged into A2A under the Linux Foundation rather than fragmenting the field. Why building on MCP plus A2A today is the safe bet.

Part 3 of the Agentic NOC series. The smallest real A2A agent: one skill, an Agent Card served at the well-known URL, and the actual message/send JSON-RPC captured live. The horizontal axis, one agent discovering and delegating to another, built from scratch, and how it composes with MCP.

Part 2 of the Agentic NOC series. The smallest real MCP server: one read-only tool, stdio transport, and the actual JSON-RPC handshake captured from a live run. Reproducible with one dependency. This is the vertical axis, an agent reaching a tool, built from scratch.

Anycast usually lets the network decide which datacenter answers. I moved that decision onto two GoBGP hosts that watch both datacenters and steer a single global VIP between them. Here's the design, the traffic flow, and three failover drills that came through with zero failed requests.

Part 1 of the Agentic NOC series. It introduces the two protocols that let AI agents reach tools and each other, MCP (the vertical axis) and A2A (the horizontal axis), and the third, ACP, that converged into A2A, plus the mental model the whole series is built on.

One HTTPS request followed from an external client to a pod, with the show commands and packet captures from every device it touches: BGP unnumbered underlay, type-2, type-3 and type-5 EVPN routes, symmetric routing proven on the wire, a Palo Alto A/A pair and Cilium.

There is a network inside a GPU server that moves more data than your entire fabric, and you will never configure it. Where NVLink ends is where your job begins: the scale-up versus scale-out boundary, and how GB200 NVL72 moved it.