Hacker News Reader

Comment by ComputerGuru | Hacker News Reader

> Assuming that CDN selection is purely geographical or route based is just as poor an assumption as your accusing others of.

As you yourself quoted, I used the word "usually." That directly and explicitly implies "not purely."

The CDN you use depends on many factors. Nearness to your route/physical location is a big factor. The relative power/bandwidth of each CDN server is another factor (weighted distribution). Internal network factors also contribute to the decision, as does maintenance, BGP peering, random load distribution, and current network health/status.

View on HN

Actually many "video cdns" rely on redirects at the HTTP level and don't work at the DNS layer at all. That's quite different to a normal CDN (where your comments do apply).

Eg, The Jetstream CDN: http://www.jet-stream.com/

I wrote my comment in response to your original comment, which was worded in a significantly harsher tone. I updated my comment with your edited quote.

Thanks for the detailed comment above. As I said earlier, I'm no networking guy! However, before posting I did test on a friend's Verizon FiOS connection, my own Verizon Mobile Hotspot, and also the various message boards I read complaining about speed appeared to be overwhelmingly TWC. I do understand that's not ideal proof. This was just intended to be a share of a quick fix.

Any suggestion to make my blog post more technically sound, I'm all ears! I can't seem to find a conclusion in all the HN comments except that "it could be a bunch of stuff going wrong", and to be honest, I don't have the expertise to find said conclusion.

The criticism was not leveled at you in particular, thank you for your willingness to revisit your assumptions. You need to test the connection to the same CDN node from a different ISP. Get the final URL for the video and try it on the other PC, and time the download for comparison. Make sure the CDN does not force a redirect to a different node.

You can get that kind of info from Chrome's "Resources" tab in the dev console (or the equivalent in whatever browser you are using). You should run the test several times to avoid warm/cold cache along the way. Should the results significantly differ, you should also note the traceroute/tracert results to the final hop.

You can try these ready-made tests from Glasnost, but they do not cover all cases: http://broadband.mpi-sws.org/transparency/bttest.php

Reply on news.ycombinator.com