Online Speedtest Results Will Vary

Recently, I worked with a customer installing a new 50M ethernet WAN circuit. This customer already had an existing 50M ethernet circuit through another carrier; however, this existing service was much more expensive. The new circuit was going to be a significant cost savings.

After a few unsuccessful attempts at turn-up (VLAN, speed, & duplex mismatches), we finally arrived at the point of expected success. That is, until the customer opened up a web browser to perform a speed test on the new circuit.

The default server the speedtest chose (based on their location) gave them 50M download, 30M upload. On the old circuit, the default server the speedtest chose consistently gave them 50M download and 50M upload. Hard to argue with what appears to be gospel to the customer.

The truth came out after we chose different speedtest servers. From the same location, using only a test laptop (directly connected to the WAN router) and no other traffic on the new line, the results were wildly different as you can see below.

Eventually, the customer was convinced that the 50M circuit was going to be fine (given the fact that 3 different servers indicated 50M for down & up speeds). We had also ruled out config mismatches, packet drops, and physical layer problems. The end-to-end circuit was very clean.

Link to download the PDF below:
https://drive.google.com/file/d/0B5ttjkGSReNeMTlBancwT093ZzA/edit?usp=sharing