Since last spring I’ve been running an on-again, off-again series about website performance. In early April I left off with some talk about hosting, server hardware, and server software, specifically Apache, NGINX, IIS, and LiteSpeed. Oddly enough, between publishing the last of those posts and writing this one, I moved the site to a different web host.
I won’t go into all the reasons for moving because they aren’t relevant. What is relevant is since I was changing hosts anyway, I thought it would be a good time to run some performance tests to compare the site before and after the move.
Today I want to talk a little bit about the move and where possible compare the specs of the new server to the specs of the old one. Then I’ll share some results of a test I ran to measure the server response times for the site before and after the move. Next week I’ll share results from the WebPageTest site and compare them to the results from the last time I ran the same tests during the previous mini-series.
Old and New Server Specs
My new host is Siteground. I’ll leave my old host unnamed as I hold no ill will toward them and don’t want to prejudice anyone against them. I chose Siteground by using a Firefox extension called BuiltWith. The extension is also available for Chrome and Edge browsers and you can test URLs using the company’s web tool.
I visited sites by writers, web designers, and web developers that always seem to be running well and used the extension to see who hosted their sites. Siteground was one of two companies that were used often and after a little research I decided to host with them.
Here’s a comparison of some of the specs between the old and new servers. You’ll notice that aside from bandwidth, the specs of the new server are uniformly better. The most significant changes are that I now have two cores instead of one and RAM has increased from 1 GB to 4 GB.
| Old Host | Siteground | |
|---|---|---|
| CPU | 1 vCPU | 2 CPU Cores |
| RAM | 1 GB | 4 GB |
| Storage | 30 GB | 40 GB |
| Bandwidth | 10 TB | 5 TB |
| Operating System | CentOS | CentOS |
| Server Software | Apache 2.2.15 | Apache 2.4.29 + NGINX |
| PHP Version | 5.6.31 | 7.2.3 |
| MySQL version | mysqlnd 5.011 dev | mysqli 5.5.32 dev |
The new CPUs each have a speed of 3 GHz and a cache of 25.6 MB. I couldn’t find the same information for the old virtual CPU, but a little searching of Google suggests the dedicated servers from the same company run at 3.3 GHz. I would assume my VPS had less, maybe 3 GHz at most, but I honestly don’t know. The new storage is all SSD where the old ones were Hybrid SSD. I couldn’t find the version of CentOS on my old host so I didn’t include versions in the table.
One thing I should note is my new server didn’t come with PHP 7.2.3 by default. Upgrading only required a handful of clicks so I upgraded the version. My old host would have upgraded for me, but I left before I had them do the work.
Also Siteground made it very easy to set up SSL over the whole site. They transferred my Let’s Encrypt certificate when they moved the site and they offer a free plugin to do the rest. As a result the site now also takes advantage of HTTP/2. My old host would have also set up SSL across the site for me, but I hadn’t asked before leaving. I wasn’t sure if they would set up HTTP/2 for me as well, though I assume they would.
I wanted to show the difference in specs between the two servers to point out the performance tests I ran aren’t exactly an apples to apples comparison. The new server on Siteground comes with better specs in general and the changes I made improved things further.
On the other hand, I should point out, that with both companies I purchased essentially the same plan or the lowest cost cloud VPS plan each offered. So while a spec to spec comparison isn’t necessarily fair, in all likelihood the plans are comparable from the perspective of someone signing up for hosting.
Server Response Time Results
I tested the server speed using the Bitcatcha Server Speed Checker. The test measures the response time for the server. They test response times from eight locations around the world, testing each location three times and using the median value as the result they display.
This test of response time should be based on the quality of the server’ network and hardware so the tweaks I made, like upgrading to PHP 7 shouldn’t affect it. Grades are assigned based on how well the speed from each location compares to Google’s recommended 200 ms as a goal, though only a single overall grade is presented instead of one for each test location. If you run the test yourself, scroll down to the bottom of the page to see what time ranges correspond to what specific grade.
The site on my old host received an overall grade of D+ and the site on Siteground received a C+ which isn’t much better. Oddly enough in the list of recommended hosts directly below the results, Siteground is listed first with an A+ rating. Here are the times from the different locations for all three.
| My Site on Old Host |
My Site on Siteground |
Siteground’s Website |
|
|---|---|---|---|
| US (W) | 64ms | 52ms | 53ms |
| US (E) | 23ms | 1ms | 1ms |
| London | 172ms | 102ms | 104ms |
| Singapore | 481ms | 425ms | 241ms |
| Sao Paulo | 526ms | 274ms | 142ms |
| Bangalore | 294ms | 327ms | 257ms |
| Sydney | 810ms | 224ms | 231ms |
| Japan | 530ms | 834ms | 143ms |
| Grade | D+ | C+ | A+ |
You might notice my site runs just as fast as Siteground’s in the U.S. (the server is located in Chicago). Where the numbers diverge is overseas and I suspect the solution is for me to set up and use a CDN. Siteground offers several free per account, which I plan to set up as soon as I get a chance, hopefully before a future series I want to write, which will include CDNs as one of the topics. My guess is I’ll set things up while working on those future posts.
Another thing to note is aside from Bangalore and Japan the site runs faster in all the cities tested on Siteground than it did on the old host. If I’m not mistaken, my site was being served from multiple servers around the world on the old host and so it’s not surprising it saw faster times in some countries.
Overall, it’s not too difficult to see why the grade jumped from a D+ to a C+ and why Siteground’s site, which I assume is being delivered over CDNs around the world, outperforms mine and gets an A+
Closing Thoughts
I’ll stop here today. Again this wasn’t a 100% apples to apples comparison as the specs on my new server are generally better across the board. I also made a couple of performance tweaks, but I don’t think they impacted this specific test and I think I may have lost a tweak or two as well in the move,
I do think it’s a fair comparison in the sense that any person who would have chosen the hosting plan I had with my old host would also have chosen the plan I now have with Siteground since both are the lowest cost cloud VPS option each company offers.
The site is running faster now, at least in regards to server response time and I wanted to share the results as a reminder that a change in web hosts, even on similar plans can potentially offer performance gains. I’d suggest it’s worth running some tests yourself against your site and the sites of several hosting companies you might consider moving to.
Next week I’ll present the results of tests I ran through the WebPageTest site along with something interesting I noticed looking at the waterfall charts that are presented with the results.
Download a free sample from my book, Design Fundamentals.