Leaderboard

Over the past few years, the Irish Three mobile network has prioritised certain types of traffic on its network. This includes the likes of Google services (e.g. YouTube), CloudFlare, some Microsoft services and certain hosts. In the past, this included traffic over certain port #s such as port 8080 that Ookla's Speedtest traffic runs over. This also made it possible to exploit the traffic prioritisaton by making a VPN connection over port 8080, however, they (and Vodafone Ireland) stopped traffic shaping by port # shortly after I posted about it. With Cloudflare recently launching a privacy App 1.1.1.1 for Android, it made me wonder if I could exploit Cloudflare's traffic prioritisation with its WARP+ service. I purchased the Unlimited subscription to try, but quickly ran into a problem. While the service worked over Wi-Fi, I could not browse the web over Three's mobile data and Apps had no connectivity. After doing some diagnostics, I found that WARP+'s DNS queries were not getting through, but TCP connectivity was working. This is surprising giving than the 1.1.1.1 App is supposed to securely encrypt DNS queries back to Cloudflare. Since connectivity by IP address was working, this gave me an idea - Enable DNS over HTTPS on Firefox for Android: Enter the address: about:config Change network.trr.bootstrapAddress to 104.16.248.249 Change network.trr.mode to 3 This gave me Internet connectivity in the Firefox App until I disconnected WARP+. I had to change network.trr.mode to 2 for Firefox to work once I disconnected WARP+. While the other Apps still had no connectivity with WARP+ enabled, I was mainly interested in running speed tests. In Ballybofey with a good 4G+ signal on Three, I ran TestMy on every server with a linear 50MB block size with WARP+ disconnected. I then enabled WARP+ and reran the tests. Here are the download speeds in Mbps: While Cloudflare's 1.1.1.1 made a substantial difference, I believe this is mainly down to Three's traffic shaping of Cloudflare traffic. It's interesting to see that even Sydney at the opposite side of the globe to me got nearly a 10x speed boost. The Los Angeles TestMy server appears to be on a server that Three prioritises as repeat tests consistently deliver around 70Mbps when even Ookla's speed test with local servers couldn't achieve anything higher than the 30s. From testing Three's masts in different areas, it appears that most (if not all) Three's 3G masts and certain 4G masts are not affected by traffic shaping. For example, in Donegal town I get much better speed with a direct connection than over 1.1.1.1 WARP+. London,GB - Direct connection vs 1.1.1.1 WARP+: Frankfurt, DE - Direct connection vs 1.1.1.1 WARP+: Sydney, AU - Direct connection vs 1.1.1.1 WARP+: I didn't have enough time to test all the servers, but it's clear that WARP+ was actually slowing down my connection in this location. Based on the London speed test, it's quite possible Three don't apply traffic shaping to masts fed directly into the fibre network. Whenever Cloudflare releases this service for the desktop, I will carry out some testing over a tethered / Hotspot connection as Android bypasses any VPN connection when tethering is enabled.

Welcome Mobycore, I think in the end you'll find that TestMy.net more closely represents your real world performance. I see that you've been testing with single thread, there's also a multithread speed test. Other speed tests usually default to multithread. In single thread mode (default) the page itself becomes the file you're downloading. In multithread a bunch of image elements are loaded. In either case the data is random, generated by the program. The transfer is timed and calculated Time/Size to arrive at the throughput. Depends on the server and the connections between your computer and the host. If your passing through a 1000 Mbps router/switch or if you're testing on one of my 1000 Mbps servers then that will limit you to 1000 Mbps. I hope to have all servers upgraded to 10Gbps in the near future. Upload has a limitation, due to the small test sizes. But a multithread upload test is in development that no longer has those limitations. If your upload speed is over 300 Mbps right now, you'll peg the current upload speed test. Different connections handle the different sizes differently. Sometimes larger sizes allow the connection to warm up and average out higher. I've seen others perform better on small bursts then get throttled. TestMy.net can help you discover odd things like this because of the odd options it has. Let me know what browser you're using and the specs of your computer. I need to look closer at both sets of results too. What did you get on LTE? Set and identifier on each device. It will help us keep better track of each one. Go to https://testmy.net/mysettings and change the identifier.

Awesome, thank you for confirming it's fixed.

I just to try out right now, and just working right now! Thank you so much too.

Think I found the problem. Can you please try again for me?