Welcome, Netflix’s new measuring service!

Welcome to the club! We at Bredbandskollen are happy that you at  fast.com have finally joined our fairly small but amazingly nice bunch of services on the internet that measure broadband. And, as always, the more the merrier.

Läs den här bloggposten på svenska.

However, as one of the old guys in this group, I thought I might pass along a bit of friendly advice. No criticism, but just little bits and pieces that we’ve picked up after many long nights and cans of copyrighted caffeinated soda, enjoy.

1. Measuring with Javascript is unfortunately not that reliable. We have spent hundreds of hours on this, but it simply doesn’t work. You’ve might have cracked it, but unfortunately it doesn’t seem so. That is why we have decided to go for native apps for iOS, Android, Windows, Mac and Linux instead of waiting for the upcoming Flash apocalypse.

2. If you run a website over https, you also have to measure with https. This takes significantly more resources in the browser and the computer, so the measurement tends to show the limitations of the computer rather than the broadband. Not a good idea. Do the measurement over http, this isn’t that secret.

3. One needs to tell the user where the server is so that the user can relay this information to their operator. And please put your server in a neutral place (for example, an internet exchange) and not with any individual operator.

4. Always have more total bandwidth on your servers than the maximum bandwidth people have on their services. In Sweden, we have some users with 1000 Mbit/s and a whole lot of 250 Mbit/s and 500 Mbit/s. This puts unreasonable demands on the servers and is a big reason why we in Stockholm now connect all of our servers with 10 Gbit/s (10000Mbit/s).

5. Uplink and response times are important for the user, of course, partly because these figures are used in marketing from the operators and partly because in principle, all services are dependent on a good uplink and short response time in order to deliver a good user experience.

6. It’s fantastic to measure over IPv6 if the user has it. However, not many websites have IPv6, so you should also supplement with a measurement of IPv4 and show both results so that the user understands what is happening.

7. And lastly, if the user doesn’t get what they pay for, ask the user to try other browsers/measuring services/computers. Ask the user to check their own equipment and make sure that the measurements, for example, are not taking place over a dodgy wifi network. Our experience is that the majority of problems are with the user and can easily be solved with some simple tips. It is just as much our responsibility to inform the user of these issues, as it is the operator’s.

Ok, I think that was everything. Once again, welcome, and good luck in the future. See you at the office party!


Tags: , , , , , , Photo: Netflix envelope by Marit & Toomas Hinnosaar (CC BY)

About the blogger

Rickard Dahlstrand Rickard Dahlstrand Project Manager for Bredbandskollen Rickard is in charge of the Broadband Check, which is used to test the speed of various broadband connections. He has previously developed similar measurement systems for testing TCP/IP networks for various companies and worked as a networking technology instructor.

Leave a comment

Reply to a comment




No comments yet.