Last week I revealed my new side project – a site that secretly runs performance tests on shared hosting accounts and reports the results. I didn’t get much feedback from you readers, but guess who did contact me – GoDaddy!
When I first got the email from GoDaddy I freaked. Were they going to be pissed that I was exposing their performance? Was I violating their terms and conditions by sending testing bots to their servers? Was my account about to be deleted – or worse – are they about to sue me?
A pleasant surprise
Actually, what followed was a very pleasant experience.
Turns out they had seen my post on Reddit about them and were curious of my testing methods. They understood I wasn’t going to tell them which server I was on (I don’t want to chance preferential treatment), but I was able to tell them what hosting package I was using, so they could help isolate the issue and work on a fix.
But then what happened shocked me – they put me in touch with another employee (the head of Hosting Research and Development) so that he could assist me testing their servers even further.
Wait – WHAT? They actually want me to be doing this?
This employee, Dave Koopman, not only works to fine tune all the GoDaddy servers and ensure they’re running fast and stable, but he’s also an industry expert in hosting performance (check out his blog). His goal is to better the hosting industry by promoting real performance reviews instead of biased opinions. He sees sites like mine as a challenge – and if I uncover some holes in their performance, he’s ecstatic it was reported and is immediately working on a fix.
GoDaddy’s response to my performance tests
Dave read through how I run my performance tests and offered some input. For the most part he thought it was actually pretty spot on:
“Shane, first I want to say, I am thoroughly impressed by your dedication to your mission. Your use of multiple tools, webpagetest, pingdom, download test, blitz.io, and support response all combined together is awesome. I have a few comments to make, and I hope you take as constructive criticism and not a bash on your method. Your method is the best I’ve seen in the wild and I’m learning from you. I hope you continue to run this and I hope we can stay in touch to learn from each other.”
– Dave at GoDaddy
After I stopped blushing I read through his extremely long, detailed email, and was very impressed. He understood every procedure I was performing, and was quite thrilled with them. He shared some test results of his own that he had performed with Pingdom, Catchpoint, and Gomez (and they weren’t too far off from mine). He also offered a few suggestions, if I was open to making my tests a little more accurate.
Note that through this all I was taking everything he was saying with a grain of salt, as he was someone that could benefit if the tests were changed to benefit GoDaddy in some way. But almost every suggestion he provided made sense and would be more accurate for all hosts, not just GoDaddy.
A few of the suggestions he provided that I particularly liked include the following:
- Page Speed Test: Do median of 9 runs, not 3 (he heard a study that said 31 is the ideal number of runs, but webpagetest.org limits you to 9). Also think about doing Native bandwidth instead of Cable (I recommend using “Native”, because the bandwidth/latency simulator is less than ideal and when you mix real latency with simulated latency, you get inconsistent results). Finally, to avoid skewing due to server location, perform one set of tests from the west coast, and another from the east coast, and average the results.
- Load Test: Consider running this from multiple IP addresses to prevent the chance that the server will go into defense mode. (Blitz.io has a setting to do this, just need to figure it out).
I really appreciate those suggestions and will be rolling them out in next month’s tests.
More than support
Just when I thought I couldn’t be surprised any more, Dave did it again: he offered to create and donate a private WebPageTest server so I could take advantage of additional testing features not offered publicly. This included the ability to run batch tests that would increase the number of tests I can run (in addition to automating the process and making it easier for me to do). He also configured it to have multiple testing locations, so I can test from the east and west coast and average the results.
Dave got this set up in less than a day and then proceeded to take time to teach me how to take configure and take advantage of the new testing features. For the nerds that are interested, the WebPageTest web server is hosted on a GoDaddy VPS, while the test agents are spun up as needed via Amazon EC2. This setup will provide no advantage to GoDaddy even though they’re hosting the web server since the testing agents are on Amazon and in very different locations.
When I use this server next month to run some tests, I will be taking a few precautions to keep the sites I test private. After all, if he set up the server, he could always find out what was happening on it. However, if he does see my list of URLs, figures out which one is hosted by GoDaddy, and attempts to game the system I’ll be able to notice a huge change in results. And if Dave doesn’t have an explanation for that, then I’ll know he had other motives (although I honestly do trust the guy, he seems very genuine).
Impressed by GoDaddy
This whole experience has really changed my view of GoDaddy. I hadn’t really used them before personally, but I’m aware the public perception isn’t the best. Regardless, I’m convinced that their team actually cares about this and is trying to change it. Did HostGator ever care when their servers were unreachable for hours? Hardly. But here’s an employee who is actively looking for problems before they’re encountered and reaching out to make things better. He’s so proud of his team and progress that he wants us to be trying to find problems!
Good job, GoDaddy. Thanks for the testing resources, and thanks for the great attitude. You’ve definitely gone up a notch in my book with your service (although they already ranked #1 in fastest support response time).