/blog
Operator notes. No marketing fluff.
Bandwidth math, panel migrations, BGP communities, postmortems from the AS209155 NOC. Written by the engineers who run the racks — not the marketing team.
How much GPU do you actually need? Sizing for AI, ML and rendering
Most teams pick a GPU on a Discord recommendation, not on data. Here is the math: VRAM sizing per model, memory-bandwidth vs compute bottlenecks, and when dedicated rental beats cloud per-hour.
How much bandwidth does a 50K-line IPTV head-end actually need?
Most IPTV operators undersize their pipe by 2-3x. The four-number formula every panel operator should know — and a worked example for a 50K-subscriber mixed EU panel.
/ subscribe
Hear about new posts.
RSS feed at /blog/rss.xml — or hop on Telegram where new posts get announced with one-line summaries.