The tale of three in-flight internet services

One of the challenges with evaluating in-flight internet services is that the comparisons are rarely head-to-head. It is not possible to compare two different providers at the same time as that would require being on two different planes at the same time and that is a skill which I don’t think anyone has quite mastered yet. Still, I came darn close a couple weeks ago, connecting from a JetBlue flight with their FlyFi Ka-band satellite service installed to a Delta flight which had Gogo‘s ATG service on board. Each of the two flights was 4+ hours giving me the opportunity to compare the performance of the service over time. And, because the flights were on the same day, I was mostly doing the same work during each, allowing me to compare real world usage conditions as close to a head-to-head situation as I believe I’ll ever get. FlyFi was the winner on this day, but Gogo put up a respectable showing.

Part of the performance metrics comes in the form of Ookla’s Speedtest.net site. These numbers are, in theory, objective and absolute comparisons of the real-time bandwidth available but my experience suggests that just looking at these numbers leaves one short on all the details. Is there a real difference to me as a user if I’m seeing 20mbit download speeds versus 10mbit or 2mbit? Does the latency matter (ATG is much lower than satellite-based services)? How does the upload speed affect the experience? Some of these are answered below.

Gogo ATG4 v FlyFi Ka-band Satellite

My test script (though it was not particularly scripted) involved a hefty mix of image uploads to social media and other services as the first flight was an inaugural and I was more or less live blogging it. The uploads carried over to the second flight as I was trying to catch up and make sure I had the details posted. And the difference between uploading on FlyFi versus Gogo was significant. Even with the higher latency the satellite-based service was better able to handle the image uploads than the ATG offer. Much of that is a function of the bandwidth allocated to uploads (a JetBlue executive informed me that FlyFi has recently increased the provisioned upload capacity on board during their beta rollout as a result of observed demand). Even where Gogo can, in theory, have similar upload speeds to what FlyFi has provisioned, neither the speed tests nor the real world trials kept pace.

I also spent a couple hours during each flight connected to a Microsoft Terminal Services session of a server on the ground. In this particular case it was a double-hop connection, connected to one server and from there to another, all over RDP. That’s probably not too common of a setup but it is what I have to do for one client and it was the same for both connections. RDP is a relatively tolerant protocol; I’ve used it over 2G cellular networks in Africa and, while slow, it was functional. It is also relatively bandwidth-friendly, especially in the manner I was using it which was not particularly graphics intensive. And for part of the Gogo flight I felt that the performance was comparable to the FlyFi I had been using just a couple hours prior. It is true that the Gogo service is not really built for this sort of data traffic – relatively steady consumption albeit not at huge bit rates – but that didn’t make my work experience any better. I was able to complete my tasks but it was notably slower using Gogo.

At one point during the Delta/Gogo flight I posited that the slower bandwidth was not really affecting my ability to get things done. And for basic services like email, browsing websites or social media both performed very well. I don’t think I would notice a difference between the two for such activities. Alas, that’s just not the way I typically use the internet, either at home or in the sky.

FlyFi test speeds — From the JetBlue Fly-Fi system

a screenshot of a phone

Panasonic Ku-band Satellite

I mentioned three services being tested in the title of this post and there was a third, though not on the same day and not under quite so similar workloads. I also had the opportunity recently to test out the international Ku-band satellite service from Panasonic on a United Airlines flight from San Francisco to Seoul. This is the second time I’ve been on a plane equipped with that kit; the first was in February on a short domestic hop. This 11+ hour trip across the Pacific was a rather different test experience. The service was, unfortunately, the least reliable of the three. At various times throughout the trip the connection would drop, even without a notable course change or other move which typically cause the on-board antenna to lose signal temporarily. There was one point where I was unable to upload email attachments even with the service supposedly active. It did clear up relatively quickly but the frequent hiccups of that that nature made for a rather frustrating user experience. When it was on the download speeds were reasonable enough and I didn’t find myself annoyed at the performance, but I also wasn’t uploading anything or using the RDP services.

I found myself waiting more than I was comfortable with for the connection to restore so that the request I was making would complete. Sometimes the outage was quick while in other cases it was multiple minutes.

a screenshot of a phone

I also had to go through the credit card authorization process twice on the flight. I slept a bit and when I went back online later in the trip the system did not recognize my laptop. The receipt number is the same for both transactions and ultimately I was only charged the one time so it did work out OK in the end, but it was mildly annoying at the time.

I fully understand that being connected to the internet while in a seat 5+ miles above the earth’s surface and moving 500+ miles/hour is an incredible confluence of technologies and I am in awe every time it works. I recognize that it is not a perfect service all the time and I am OK with that, even if I get annoyed when it isn’t exactly how I think it should be. At the same time, getting to compare the various technologies directly is not all that common an experience and there are real differences between them. As the industry matures it is these differences which will help define the winners and losers in the market.