Tracing down internet connection speed problem

[cstech]

I have a client that wants some kind of definitive answer and I don't know how to give one. The problem; he has 100Mbps internet connection speed from the local cable company. According to them he is getting the speed he is paying for (twice the cost of 60Mbps) However when we do a speed test it fluctuates greatly in actual speed. He primary used his iPhone 6 and other devices in the house are all updated, mostly apple products. He has a home automation and lighting system (Elan G!, and Lutron RA2 for lights) that is also connected to his network. He is using a Apple AirPort Extreme with the TimeCapsule as his main router and he really doesn't want to change this. We have turned off the wireless in this device and changed out the other AirPorts he had with Ubiquiti AC-Lites. Part of the problem he was initially having was that he has a large house and one wireless was not enough however he was still close enough that devices would not let go of the AirPort downstairs and connect to a different stronger signal. We now have Two AC-lites in the house with the antenna's dialed down. The controller is saying that devices are connecting to the right APs according to their location and getting very good signal strength on each device. Most devices are connecting between 72Mbps and 300Mbps. Nothing really below that. We have also tried on hardwired devices that register as full duplex gigabit speeds connecting directly to the router yet the download speed fluctuates there as well.

What are some other causes that could have his speed when running a speedtest.net test go between 95Mbps on the iPhone to 23Mbps on the same phone just a few hours later? That is too big of a jump to be neighboring traffic on the cable line as well as the fact that he is one of few on the node. (Think mountain area with scattered houses, and I checked with the cable company to see, according to them that node is only half full)

 

[lcoughey]

What happens when you test the speed when connected directly to the main router, with everything else disconnected? I also wouldn't trust a wireless device, especially a mobile phone to get consistent results either.

 

[cstech]

Connected to the main router I still get fluctuations but not as "severe" but it is very intermittent still.

And I agree with not trusting wireless devices for speed tests however that is what the client is using to "verify" his connection. Somebody will come over and say oh your internet is slow and he will run a speedtest from his phone and say oh yeah I'm only getting 24Mbps instead of the 100 I am paying for and send us a screen shot of the phone test and say "You'r system is not working, fix it now" I just need something that can explain to him what may be happening that he can understand.

 

[YeOldeStonecat]

Becomes very time intensive....first thing I'll tell a client is "broadband is not guaranteed bandwidth! If you need guaranteed internet...get a biz grade ethernet or fiber connection!" Now..obviously paying 600+ bucks per month for a biz grade dedicated pipe isn't in most residential users budget...or even many SMBs budget. But..that's the reality. YOU, the computer tech, are not responsible for the connection between the house..and the ISP. Only thing you can control is their router, and their wired devices. Notice I left out the wireless? That's something you can't really control either!

Like mentioned above..disconnect EVERYTHING from the router...and start running an online speed test while plugged into the modem/gateway..with a wired computer. Run many tests throughout the day..different time periods...for several days. Now you have your baseline.

Next...connect the router back up...but disable the wireless...and plug in just 1x wired computer...and run speed tests. Comparable speeds to being just plugged into the modem/gateway? Should be...but if not, you know the router is your potential bottleneck....or at least 1 of the potential bottlenecks.

Do not run speed tests using wireless! It's not 100% consistent.

Keep running speed tests using WIRED computers....that will be a more true reading.

If wired is solid and stable...and wireless is the varying one...now you can start tuning his wireless a little bit. Home automation...is that equipment "phoning home" all the time? Or hourly throughout the day? What else for wireless clients does he have...that may be constantly "phoning home" to something?

How many devices does he have total on his wireless? Can you get a dual band AP in there and split up his load? Because don't forget...each wireless client divides wireless bandwidth/speeds by +1. Since wireless is not full duplex.

Do a wireless site survey...granted he may not have much neighboring wireless networks....but within his own house he may have a lot that interferes with wireless. The 2.4 frequency may be saturated...check those channels, split them, make sure they're not overlapping. And if he has 5.0 clients...setup dual radio APs and have a 5.0 SSID for any 5-oh clients he has. I just scrolled up and noticed you have AC-Lite APs..so that should do it...make sure some clients are on the 5.0.

 

[ComputerRepairTech]

and dont rely on a single speedtest.net server, try others.

 

[phaZed]

Ping and Tracert and make sure your getting good times on those.

 

[Metanis]

I hope you are charging by the hour for this. It will be a nice paid education!

 

[Markverhyden]

As has been mentioned. Do not rely on one type of tool for speed tests and understand that the speed SLA is not what the cable operator advertises but what is in the small print. They almost always say something like "up to" or "maximum". And I'd bet their small print also says something about being hard wired to the router, single device, etc, etc.

This is some of the things I do when dealing with internet speed issues. Web based speed tests on speedtest.net, speakeasy.net, and internetfrog.com. I'll run one after the other. Then I'll load all three into separate browser tabs and run them together. I'll also run CLI continuous pings for several minutes to both an IP as well as a FQDN. Remember that to process the FQDN request DNS must be involved.

Another tool I have used in the past is iperf. It's a great tool for understanding the underlying pipe and can be used to demonstrate the difference between pipe size and actual user experience. So many user think that because they doubled their pipe that the internet performance will double for them. Not so.

 

[altrenda]

The Airport Extreme with Time Capsule (or without) is a POS. Designed for simple setup on Apple machines by end users, not for performance.

have you upgraded the firmware on the Airport Extreme?

Have you tried subbing in a different router and running some tests? Even if the client wants to keep the Airport Extreme, at least you know.

 

[dnt button]

Is this company of any help? https://www.samknows.com/

 

[fencepost]

If he's using wireless for checking speed, then this may be worth watching:

I watched it after someone posted it here a couple weeks ago, very informative on some of the things that can cause WiFi issues.

In particular are there any other wireless devices in the house? This sounds like the kind of person who will have bought all the hip new wireless gadgets including cameras, cordless phones, baby monitors, this that and the other. All of those can cause interference that'll destroy your speed.

 

[Markverhyden]

If he's using wireless for checking speed, then this may be worth watching:

I watched it after someone posted it here a couple weeks ago, very informative on some of the things that can cause WiFi issues.

In particular are there any other wireless devices in the house? This sounds like the kind of person who will have bought all the hip new wireless gadgets including cameras, cordless phones, baby monitors, this that and the other. All of those can cause interference that'll destroy your speed.

I've got a Wi-Spy and it's a really nice tool.