I had dabbled with CDN before (LiteSpeed, Jetpack), but a recent requirement around my NAS required me to move my DNS service to Cloudflare. So I thought, why not use their CDN, too and compare the before and after performances?
For web traffic, you can use Cloudflare in two modes
- DNS only
- DNS plus Proxy
As the names say, the first option uses Cloudflare servers as your authoritative name servers, but your traffic goes straight to your original server. There’s no caching and there’s no CDN.
Turning on the proxy is where the real benefits are supposed to be visible. I used their free plan for testing.
As you can see, I used various tools to test the difference in performance.
Network Latency
Being a network engineer, the first thing I tested was network latency difference between my web server and Cloudflare’s proxy.
I tested from Thailand, and as expected, Cloudflare’s proxy has significantly lower latency compared to my web server, which is based in the US. The benefits may be lower if tested from the US itself.
Average 224ms latency vs 36ms.
This means that my users can experience better Round Trip Times to my website globally by connecting to their nearest CDN, instead of all focusing on one server in the US.
But ping is only part of the story. For real-life results, I used the below tools.
I ran all the tests raw with Cloudflare paused and WordPress plugin disabled. Then I re-ran the tests again, one day after enabling both.
Google PageSpeed Insights
Google PageSpeed Insights is an important tool, especially because Google uses this score as a metric to calculate its search engine Page Rankings. It is a bit simplistic, as in it doesn’t let you select a source server. But it does show results for both mobile and desktop.
As you can see, there’s a minimal performance gain, which makes sense because Google’s servers are likely in the US, similar to mine.
DebugBear
DebugBear is similar to Google, but also shows a helpful graph of the different activities happening during your page load. Again, only US is available as a location for free.
As you can see, there’s an insignificant decrease in the load score for mobile, although desktop being perfect to begin with couldn’t be improved on.
Solarwinds Pingdom
Pingdom has an advantage over the last 2 tools in that it allows you to select a server location to test from. I chose Australia, being far from USA.
As you can see, there’s an insignificant amount of decrease in Pingdom’s score, but the load time and the number of requests are down significantly.
Wattspeed
Wattspeed’s lighthouse tool allows you to test for Desktop and Mobile separately, but doesn’t let you select a source server.
Significant increase in scores for both Desktop and Mobile.
Overall Verdict
| Test Tool | Raw Results | Cloudflare Results |
|---|---|---|
| Google PageSpeed (M) | 80/100 | 84/100 |
| Google PageSpeed (D) | 99/100 | 99/100 |
| DebugBear (M) | 71/100 | 91/100 |
| DebugBear (D) | 100/100 | 100/100 |
| Pingdom | 75/100 | 74/100 |
| Wattspeed (M) | 82/100 | 99/100 |
| Wattspeed (D) | 76/100 | 84/100 |
The verdict is clear, in most cases, Cloudflare provides significant and measure-able improvements in page load times. Considering that the service is absolutely free, it is a no-brainer to use it.
However, performance benefits are not the only advantages
- Cloudflare will stop DDoS attacks on your website.
- Cloudflare allows you to set up firewall rules to block certain traffic to your website (eg on the basis of geolocation).
- Even if your website ever goes down, Cloudflare can continue exposing your website from its cache.