Optimizing TCP for high WAN throughput while preserving low latency (blog.cloudflare.com)

Very very interesting read, I learned a lot. We recently had a problem where default configured linux system on a 180ms latency connection (east coast to finland) could only push 4mbps stable. Forcing SO_RCVBUF on listening socket of app to be 16MB kind of solved it but it was not perfect.

Question. What is the extra CPU overhead of always checking for collapse once the queue exceeds 6MiB?

Any chance at getting all the patches here upstream? https://github.com/cloudflare/linux/commits/master

Good but long read. I think it'd take additional time in creating the articles, but any plans to make short videos on such blog posts so that's easier to consume ?