Australian internet… what is going on with you, Telstra?
I have a nice 30 meg cable internet connection with Telstra. Here’s what Speedtest.net looks like for it…
It’s a rubbish upload speed — really rubbish — but the download is pretty nice. So, I tried downloading some files the other day from Flickr, using my Telstra internet connection. Despite using a wired connection directly into my router, it was going really, really slowly: nowhere near the 30Mb/s I ought to be able to get. I wonder why, I thought. So I dug a little further.
Here’s a graph, courtesy of the PeakHour app, showing what’s going on at my router:
On the left: I’m downloading a large ZIP file (about 120GB) using my standard Telstra connection. Downloads are in green. Then there’s a pause while I turn on my TunnelBear VPN (set to the US), and then I download the same file again. Can you spot the difference? While the TunnelBear VPN delivered file is a lot more bursty, it is two or three times faster than Telstra.
VPNs shouldn’t work like this. VPNs should slow down your data (you’re putting a load of extra hops into the journey, plus a whole extra set of encryption). They shouldn’t ever speed up your connection. Should they?
What’s going on?
The Flickr files are one big .zip file, served via https from https://downloadr.flickr.com/ so I wondered whether the https encryption was making a difference. So, I made a 100MB file of random data with the following terminal command:
dd if=/dev/random of=testfile.txt bs=1024 count=102400
…and then zipped it so my browser would download it, rather than try to display it. I then uploaded it somewhere outside of Australia (in this case, Dublin Ireland), in both http and https formats. So, we’ve a file which can’t be compressed further, available in non-secure and secure formats.
I then proceeded to try to download it, using
curl -o /dev/null http://s3-eu-west-1.amazonaws.com/static.media.info/testfile.zip
And the results are…
With Telstra:
https: average download speed 696k; time spent 2'27"
http: average download speed 691k; time spent 2'28"
With TunnelBear US VPN on top of Telstra:
https: average download speed 1311k; time spent 1'18"
http: average download speed 1215k; time spent 1'24"
Given that the file is in Dublin, it should clearly fly along if I was to use the local Irish TunnelBear VPN, so here comes another test…
With TunnelBear Ireland VPN on top of Telstra:
https: average download speed 391k; time spent 4'21"
http: average download speed 426k; time spent 4'00"
Hmm, well, that’s a bit strange, too.
But at least we’ve learnt that https or http appears to make little difference. There’s nothing strange going on with getting an encrypted version. And, once more, the TunnelBear US VPN has doubled the speed of this download. Even when the file isn’t in the US.
I don’t really understand what’s going on here, but the only explanation I can really come up with is that Telstra is doing something strange to my downloads that I can bypass if I use a VPN.
So, fellow internauts: is there something else I can be doing to understand what’s going on here? Do let me know in the comments, and I’ll share the response: because if it’s happening with me, it’ll be happening with someone else, too…
Via Facebook, Ian Deeley asks whether I can do a Speedtest run to a server that isn’t next door. So, here’s a Speedtest run to Dublin, Ireland (without a VPN). It ought to be the same as the 696kb/s that I managed above (er — is that right?)…
Well, that’s strange.
Via Twitter, @craigwebster asks “What does a traceroute look like off the VPN and on?”, and so here is the results to downloadr.flickr.com from Telstra…
…and the equivalent to my test file…
Hmm. I don’t think these are very helpful.
