Thai ISP performance just gets stranger. I verified today that using speedtest.net without involving a proxy server is way slower than using a proxy server. Throughout this afternoon I tested it three times, and the results are all similar:

Using "Hotspot Shield" by AnchorFree (http://anchorfree.com) I got this typical result today:

Disconnecting from the VPN, I get:

I originally was "forced" into using a VPN because a program I use to interface with my employer's mainframe computer apparently uses some port number that is blocked by my ISP (TT&T). I just accepted that as a "mai bpen rai" thing, but then noticed that if I left the VPN connected, my web surfing was more stable: it seemed faster and less sites failed to open.

Then today, I decided to test VPN vs non-VPN with speedtest.net, and sure enough, it's faster using the VPN. Any ideas on why that would be? I would have expected the opposite, as packets need to be routed through a longer path, no?

If your VPN is connecting to a workplace within Thailand, it may simply be a case that your employer has more international bandwidth available than you do at home (domestic internet speeds in Thailand are actually quite fast, so they won't normally pose a bottleneck). It may also be the case that your home ISP is throttling certain ports or kinds of traffic. You can't tell what kind of traffic is running through a VPN connection, so your ISP can't discriminate.

And occasionally a longer path is faster - back when I had a lousy connection with True I found that I could consistently access sites in the US faster connecting via an Australian proxy than I could connecting direct. I presume that ISP routing isn't always 'optimal'.

That really seems to be a false speed test as only one (Maxnet) is from Thailand. The other appears to be from the VPN server in the US to LA. Not an indication of what you would get using it from here. And a pretty lousy circuit overall. Try doing a real download and see what speed it comes in at. That is a very sick looking Maxnet circuit.

The other appears to be from the VPN server in the US to LA. Not an indication of what you would get using it from here. And a pretty lousy circuit overall.

Ah, that's probably what it is: the speed between the VPN server (in the eastern USA) to L.A. And, yes, "a pretty lousy circuit overall" seems to define most users' experience with TT&T these days.

If your VPN is connecting to a workplace within Thailand, it may simply be a case that your employer has more international bandwidth available than you do at home (domestic internet speeds in Thailand are actually quite fast, so they won't normally pose a bottleneck). It may also be the case that your home ISP is throttling certain ports or kinds of traffic. You can't tell what kind of traffic is running through a VPN connection, so your ISP can't discriminate.

And occasionally a longer path is faster - back when I had a lousy connection with True I found that I could consistently access sites in the US faster connecting via an Australian proxy than I could connecting direct. I presume that ISP routing isn't always 'optimal'.

Both the VPN server and my employer's mainframe are in the USA. I pretty much assumed that some port that was being blocked locally by TT&T (like they do to attempt blocking of VoIP and torrents) was the problem, so I tested using a VPN where traffic is not "screenable" by TT&T.

Even though I tend to agree with lopburi3 that the speedtest.net result is between the VPN server and LAX, subjectively it seems that browsing the internet is faster using the VPN. <snip>