Exploring Bandwidth Issues

I've had AT&T as my ISP since 2005.  First, it was DSL, and recently, their U-verse brand of service.  Based on the Wikipedia entry, it is either a fiber optic connection or DSL, depending on the area.  My U-verse plan advertises up to 18Mbps download speeds.

With the explosion of bandwidth-hogging apps like YouTube, Skype, even Facebook, and multiple devices sharing the same connection, I have become more conscious of the actual download speeds I am experiencing.  Even my 2 year old child starts complaining when the high-def video she is watching on TV via WiFi starts to freeze while the TV attempts to buffer more data.

AT&T recommends ookla.com to run speed tests from your computer.  They also prescribe using a wired connection (Ethernet) when running the test, as well as ensuring that no other devices are using the connection.  So with that, here are my results.  With a promised 18Mbps download speed, I am actually getting 22Mbps on a wired connection.  That's not bad.

However, as with most homes these days, devices aren't tethered to an Ethernet cable.  I use WiFi almost exclusively - 2 laptops, 3 tablets, a TV, a printer (although I doubt that needs to connect to the Internet).  So running the same test, except via WiFi, on the same computer, with all other devices blocked, I get this, 6.2Mbps.  That's less than a third of the original 22Mbps, or a third of the promised "up to 18Mbps."

I can understand some loss due to the need to go over the air vs. Ethernet, but that drop is too drastic.  A quick Google search tells me my problem is not unique.  There are several suggestions, like ensuring all the devices connecting are on 802.11g at least.  I double-checked the laptop I was using, and it is indeed capable of 802.11g, and with that device using the connection exclusively, 6.2Mbps is too low.  Finally, I was testing with a computer that was about a foot away from the wireless router, so signal degradation should not be a factor.

Another suggestion was to play around with the WiFi channels.  Apparently, if my neighbor was on the same channel, we will cause interference on each others' networks.  So switching my channel to another one, I ran another test, and I get 10.7Mbps - a good improvement, but not great.  

I then became curious as to whether the added WPA2 security decreased performance, so I turned off security and ran another test while on the same channel as earlier, and recorded 9.39Mbps download speed.  So maybe the security settings don't decrease performance.  Besides, turning off security is a bad idea, unless you really intend to share your WiFi with your neighbors.

So maybe it's time of day.  It is 5:03AM PDT, and my download speed is back to down to 6.43Mbps.  The earlier tests were done around 9PM Pacific.  Running the same test a few minutes later yields even worse results, at 2.3Mbps.  So maybe it's not my connection, the whole Internet is busy!  (Well technically it could be only the network around me that are busy.)

It's times like this that I wonder whether switching to Comcast will improve things.  But a quick Google search turns up similar complaints from Comcast users.  So maybe the whole world is hungry for bandwidth, which is why companies like Infinera and Juniper are setting the stage for really fast networks.

But back to my problem, I still need to find an explanation for the loss of bandwidth.  18Mbps speeds are achieved on 100Mbps Ethernet.  Why only 10Mbps on 54Mbps 802.11g wireless, even if it is the only device connected to the wifi router?  It might be time to go back to my Networking 101 class.