Does Cloudflare increase Time To First Byte (TTFB)? Tested on 3 hosts. 6480 manual measurements

Share the knowledge...

Facebook

Twitter

Linkedin

One of my readers contacted me regarding experiencing bad Time To First Byte (TTFB) on his website. After a little research I suspected that using Cloudflare and/or a caching plugin can influence TTFB. And I decided to make thousands of speed and TTFB tests with and without using Cloudflare as well as with and without using a caching plugin. I tested it on three different shared hosts.

I did not know what results I will get. So I was very excited to do this research. And you are welcome to see my finding s below! A lot of charts ahead 🙂 Don’t worry, I explain everything!

For your convenience: here’s the link to the conclusions.

How I tested the influence of Cloudflare on Time To First Byte (TTFB)

Here are the special things in this research:

• Smallt WordPress sites tested. Since I wanted to focus on Cloudflare and TTFB relationship, using small WordPress websites for tests was enough. The tested websites were practically identical.
• Repeat views matter! I paid attention to how Cloudflare and caching influenced TTFB not only during the first visit, but also how TTFB (and speed) changes for repeat views. In other words, I differentiated the test results by imitating visitors not only first opening the website, but also browsing the website.
• Testing locations closest to the hosting servers. Before all I wanted to check how Cloudflare influences TTFB (or server response time) and ignore network latency as much as possible. Therefore I tested speed and TTFB from the locations which are as close as possible to the hosting servers. SiteGround and GreenGeeks were tested from Chicago (from the same city where hosting servers are located). HostWinds’s server is in Seattle, and I tested it from San Francisco.
• How Cloudflare affects speed of lightweight pages. Although the major aim of this research was to analyze how Cloudflare affects TTFB, a side effect of the research was the influence of Cloudflare on small website speed.
• Cloudflare can affect differently on different hosts. Also, a side effect of the research was a finding that using Cloudflare can have different consequences on different hosts.

And here’s the procedure how I tested the TTFB and speed with and without Cloudflare and a caching plugin:

1. Installing WordPress, Twenty Seventeen theme on a shared hosting (the cheapest plan is used).
2. Installing BulletProof Security plugin.
3. Running 15 tests. Each test is 9 runs. Each run is the first view and repeat view. Each view contains data on Full Page Load Time and Time To First Byte (TTFB). The testing tool is webpagetest.org (Here’s the article why I use this tool). And this article explains why multiple tests are required.
4. Installing a caching plugin. I used a free WP Super Cache in this research (by the way, Here’s my research where I compare a couple of caching solutions). And re-running the 15 speed and TTFB tests the same way. For is SiteGround though I used the free plugin (SG Optimizer) which the host has developed specifically for using on their servers.
5. Adding the website to Cloudflare (free plan), enabling Auto Minify of CSS, JS and HTML, Cache-Everything rule is not enabled. And re-running the 15 tests. Thus, both Cloudflare and a caching plugin are used.
6. Removing the caching plugin (with Cloudflare still enabled) and re-running the 15 tests.
7. Thus, all four combinations (with and without Cloudflare, with and without caching plugin) have been tested on one hosting. And then doing the same tests for the other two shared hosts (the cheapest plans).
8. Putting together speed and TTFB data in tables and in charts, determining peak values, calculating averages. There are 6,480 measurements in total. By the way, all tests (raw data in HTML formal) you can download here (125 MB).

The most general results – How Cloudflare affects website TTFB and speed

As you will further see, Cloudflare has different effects on TTFB and speed on different hosts. But the general pattern stays the same. And here we go:

And the same data in another perspective for your convenience:

The charts data does not contain the peak values (I analyze single peaks occurrences further). The peaks are the occurrences when the value jump by more than 30% after the value series is sorted. But even if the peaks are included in the charts data, it does not change these summary charts in a significant way. And the conclusions are the same.

What I can say looking at the summary of the tests:

• In general, adding Cloudflare slightly increases TTFB if no caching plugin is used. And adding Cloudflare does not affect TTFB if a caching plugins is already used.
• The same effect Cloudflare has on website speed, i.e. if no caching plugin is used, then Cloudflare actually slightly slows the website down. Using a caching plugin helps to avoid this negative effect. By the way, similar results (when Cloudflare slows a website down for some locations) I got in another research in which I tested multiple locations.
• Finally, using a caching plugin and not using Cloudflare appears like the best solution. Actually this is not as surprising as it may seem. Especially if visitors are close to the hosting server. At the same time, Cloudflare’s cache-everything rule, as my other research shows, is a definitely a better option if you can use it.

General results per host – How Cloudflare affects website TTFB and speed

SiteGround: General test results of how Cloudflare changes TTFB and website speed

Here’s the summary of the tests for SiteGround:

As you can see, in general, there’s very little difference between different configurations in general. It regards both TTFB and website speed. And although these results may surprise you (like, hey, why caching does not help at all?!), here’s a simple explanation.

SiteGround already uses a caching solution on their servers by default. And since my testing website is small and the tests run from a very close location (both hosting server and the testing location are in the same city), there’s no visible effect of Cloudflare (or of the caching plugin) on website performance.

At the same time, the chart above does not include peak data. And below you will see that Cloudflare (as well as caching plugin) do influence TTFB (and website speed). I will analyze it further. But even on average charts this is noticeable (look at the animated image below):

As you can see on the animation above, the most stable solution does not include Cloudflare (and includes the caching plugin from SiteGround). Further I will show more detailed charts with Cloudflare causing peaks.

HostWinds: General test results of how Cloudflare changes TTFB and website speed

Here’s the summary of the tests for HostWinds:

The difference ofTTFB values between using Cloudflare and not using Cloudflare is not quite visible. But what really matters is using a caching plugin. TTFB is pretty less when a caching plugin is used. By the way, caching plugin is better for speed as well than Cloudflare (especially it matters for repeat view speed).

The charts for HostWinds and Siteground let us reveal how differently the effect of using Cloudflare (as well as caching plugins) can be depending on a hosting.

And let’s have a look at the difference of the charts with and without peaks values included in calculating averages (animated chart is below):

As you can see, the peak values don’t add up much to the averages. But still, you can see that Cloudflare causes (or at least does not help to cope with) peaks occurrences. Whereas a caching plugin gives the most stable results (see how animation does not move for “No Cloudflare – with caching plugin” section).

GreenGeeks: General test results of how Cloudflare changes TTFB and website speed

The last testing host – GreenGeeks. And here’s the summary of the tests:

Although TTFB is a bit different for the tested configurations, it does not change a lot depending on whether Cloudflare is used or not. But the interesting part is that using Cloudflare makes a website noticeably slower compared to not using Cloudflare. It’s true for both first and repeat views.

Now let’s see a similar chart with peak occurrences included into the average values displayed (the animated chart below):

Unlike other hosts, Cloudflare on GreenGeeks showed the most stable results (lesser peaks).

Intermediate conclusion of how Cloudflare changes TTFB and website speed

Here’s an animated chart to see all hosting performance with and without peaks:

Analyzing the average TTFB and speed values after running 2160 tests on each of the three hosts we can draw the following conclusions:

1. Cloudflare has difference effect on different hosts. The less effects Cloudflare had on SiteGround hosting (which was the fastest in terms of speed).
2. Smaller TTFB does not mean faster host (e.g. SiteGround had the worst TTFB, but the best speed).
3. In many cases TTFB (as well as speed) becomes worse when using Cloudflare (but not dramatically worse though).
4. No surprise if using Cloudflare causes more peak occurrences (significantly slower TTFB or/and speed compared to the rest of the tests).
5. Caching plugin in many ways looks like a better solution (for both TTFB and speed).

This was a general overview. Now let’s dive in into testing results on each hosting.

Detailed results how Cloudflare changes TTFB and speed

They say “the devil is in the detail”. So, the deepest understanding of how Cloudflare affects TTFB (and speed) you can get out of my research for each host.

Here we go.

SiteGround: Detailed results how Cloudflare changes TTFB and speed

Chronological observations

Here are four charts of observations displayed in the chronological order of tests (i.e. no special sorting). Each chart displays a testing configuration (with or without Cloudflare and with or without a caching plugin).

The chart above represents pure hosting performance – no Cloudflare and no caching plugin are used. It looks fast, pretty stable with a couple of occasional peaks (both speed and TTFB). Anyway, I’ll analyze stability and peaks a bit further.

Okay, added now a caching plugin (the chart is above). And the results are neater.

Removed caching plugin and added Cloudflare (see the above chart). Both TTFB and speed got a few peaks. Now it looks even a little bit worse than without Cloudflare.

Adding a caching plugin additionally to Cloudflare does not change things much (see the chart above).

And now all the chart in one place (for convenience):

Not a very big difference in fact.

Now let’s see basically the same TTFB and website speed data but from a different perspective.

Sorted series

Let’s sort the series by value to see the distribution of values.

What we see here? When a visitor who is located close to the hosting server first opens a small website hosted on SiteGround, there’s almost no difference whether the website uses Cloudflare or caching plugin. The speed is more or less the same at different website configurations. Only a small portion of the tests gave a bit different results (right-hand side of the chart)

Time To First Byte at the first visit is very stable and practically the same regardless of whether you use Cloudflare or a caching plugin. A slightly worse performance a combination “with Cloudflare and no caching plugin” shows. But the difference in fact is insignificant.

Repeat View is even more stable than the First View. And Cloudflare adds up tiny portion of instability (tails going up at the right side). But in general, no big difference really.

Time To First Byte is also almost identical in all four website configurations. Although a slight negative influence of Cloudflare is more visible.

And here’s all four charts in one place for easier comparison:

Sorted series without peak values

For better view on how Cloudflare (and a caching plugin) influence website performance including TTFB, I remove the peak values. Here are the result charts.

As you can see, in general, there’s no big difference in website configuration. But a little bit better speed values for the first view are achieved without caching plugin.

Here’s generally the same situation as with the chart above. I.e. no significant difference in TTFB. But Cloudflare and a caching plugin give a little bit worse TTFB at the first view.

The chart above with peaks excluded demonstrates that tests give very similar results for whatever website configuration.

Time To First Byte at repeat view demonstrates that the website configuration with Cloudflare give a little bit worse results.

Analyzing peaks

The speed peaks at SiteGround were less than 1% of all tests. Also, what is interesting is that a caching plugin with no Cloudflare gave no speed peaks at the first view.

More Time To First Byte peaks at the first view were with the configuration when both Cloudflare and a caching plugin were used. There was a small number of peaks anyway.

A pair of speed peaks at repeat views were observed only when using Cloudflare. Not a serious issue though.

There are more TTFB peak values at repeat view. No peaks noticed for the configuration without Cloudflare or caching plugin. The configuration with Cloudflare and no caching plugin gave more TTFB peaks than other configurations.

Here’s a summary chart and table displaying the number of peak occurrences:

As you can see, the configurations without Cloudflare (with or without a caching plugin) gave more stable results (less peaks).

The rest of the charts with tables in this section are the derivatives from the above table data for SiteGround. They will let to analyze the influence of Cloudflare and a caching plugin on TTFB and speed from particular perspectives.

As you can see, if looking from a general perspective, there’s almost no difference between using or not using a caching plugin. TTFB and Speed peaks are summarized, as well as summarized the First view and repeat view peaks.

If accumulating the peak occurrences with and without a caching plugin (all: TTFB, speed, as well as first and repeat views), the difference becomes more vivid. Cloudflare causes more peaks.

If the tests are not differentiated by using or not using Cloudflare, then the there’s almost no difference between using or not a caching plugin.

Let’s dive in a little deeper now by looking at Time To First Byte peaks separately. As you can see, now the difference is more visible. Using Cloudflare causes more TTFB peaks in general.

The summary chart and the table demonstrate it as well.

This chart shows that when a visitor loads a webpage the second (repeat) time, the risk of peak occurrence is higher.

Intermediate conclusions for SiteGround

When testing SiteGround’s Time To First Byte and speed at different website configurations (with and without Cloudflare and/or caching plugin) the difference between the configurations was not dramatic. Although the number of peaks demonstrated that using Cloudflare causes more peak occurrences. Whereas a caching plugin was a positive factor compared to Cloudflare.

Thus, if you are using SiteGround, then using a caching plugin is highly recommended. And using a Cloudflare (a free plan, Cache-Everything rule disabled) can be ambiguous.

As regards TTFB values, Cloudflare does not affect it in a significant way in general.

HostWinds: Detailed results how Cloudflare changes TTFB and speed

Chronological observations

The following four charts display the test results in the chronological order (i.e. no special sorting). Each chart displays a testing configuration (with or without Cloudflare and with or without a caching plugin).

The chart above represents pure hosting performance – no Cloudflare and no caching plugin are used. It looks very stable with a number of insignificant peaks (both speed and TTFB). Also, I’ll analyze stability and peaks a bit further.

Added a caching plugin (the chart is above). The results are better.

Using Cloudflare (see the above chart). Both TTFB and speed got a few peaks. Now it looks even a little bit worse than without Cloudflare.

Adding a caching plugin additionally to Cloudflare improves the stability and speed to a great extent compared to just using Cloudflare (see the chart above).

And now all the chart in one place (just for convenience):

The difference is pretty noticeable.

Now let’s see basically the same TTFB and website speed data but from a different perspective.

Sorted series

Let’s sort the series by value to see the distribution of values.

What we’ve got here? When a visitor who is located close to the hosting server first opens a small website hosted on HostWinds, there’s a steady difference between website configurations. Cloudflare and caching plugin give the best speed results. Then goes the variant with using just a caching plugin. And using Cloudflare is better than not using anything at all.

Also it’s interesting that a caching plugin saves from a little speed decrease (a jump on the chart from the right-hand side). Whereas Cloudflare alone does not help.

As you can see above, Time To First Byte is comparatively better when using a caching plugin. Moreover, there are several TTFB peaks which also take place with CloudFlare but not with a caching plugin.

Website speed at repeat view is considerably better when a caching plugin is used (with or without Cloudflare). Also, using Cloudflare or/and a caching plugin helped to avoid a number of peaks.

Time To First Byte actually repeats the pattern of Full Page Load Time (speed) chart. Cloudflare does not help much to improve TTFB, unless it’s used with a caching plugin.

And here’s all four charts in one place for easier comparison:

Sorted series without peak values

For better view on how Cloudflare (and a caching plugin) influence website performance including TTFB, I remove the peak values. Here are the result charts.

It’s the same chart as without peaks. The website configurations with a caching plugin were better in the tests.

Without TTFB peaks the situation is basically the same as with the peaks. I.e. TTFB for the websites which use a caching plugin is much better. At the same time, TTFB below 0.6 seconds is still fine.

After removing speed peaks at repeat views the chart does not bring in any new information. Website speed is better with a caching plugin, and Cloudflare does not influence much.

And TTFB without peaks looks similar to what we have seen above. Better TTFB at repeat view is like at the first view, i.e. with a caching plugin. And Cloudflare does not affect much.

Analyzing peaks

Full Page Load Time at the First View does not have peaks. I.e. no sharp (more than 30%) increase in the ordered series.

Let’s move on to the next chart.

As you can see, only the configurations with a caching plugin don’t have peaks. And the TTFB peaks are even higher if Cloudflare is used.

There are not many speed peaks at repeat views. The most serious is just one test which happened when using Cloudflare. The two other occurrences are not really high. Not bad speed at repeat views after all.

It’s noticeable that TTFB at repeat view is worse than TTFB at the first view. Also, as you can see the repeat view TTFB (the chart above) has bad luck with Cloudflare and no caching plugin.

Here’s a summary chart and table displaying the number of peak occurrences:

As you can see, the configurations without Cloudflare (regardless of with or without a caching plugin) gave more stable results (less peaks).

The rest of the charts with tables in this section are the derivatives from the above table data for HostWinds hosting. They will let to analyze the influence of Cloudflare and a caching plugin on TTFB and speed from particular perspectives.

As you can see, if looking from a general perspective, there’s a big difference between using or not using a caching plugin. TTFB and Speed peaks are summarized, as well as summarized First view and repeat view peaks presented in this chart and the table.

If accumulating the peak occurrences with and without a caching plugin (all: TTFB, speed, as well as first and repeat views), the difference is also vivid. Cloudflare causes more peaks.

If the tests are not differentiated by using or not using Cloudflare, then we can see that there’s also a big difference between using and not a caching plugin. Using a caching plugin is much better for avoiding peaks.

Let’s dive in a little deeper now by looking at Time To First Byte peaks separately. As you can see, now the difference is also visible. Using Cloudflare causes more TTFB peaks in general, but Cloudflare also allows decreasing the number of peaks at repeat views.

The summary chart above demonstrates not very dramatic difference between a number of peaks in case of using or not using Cloudflare.

This chart above shows that when a visitor loads a webpage the second (repeat) time, the risk of peak occurrence is higher.

Intermediate conclusions for HostWinds

When testing HostWinds’ Time To First Byte and speed at different website configurations (with and without Cloudflare and/or caching plugin) the difference between the configurations was pretty dramatic. The number of peaks demonstrated that using Cloudflare caused significantly more peak occurrences.

Using a caching plugin was a positive factor compared to using Cloudflare. Using Cloudflre with a caching plugin is also an advantageous combination.

In short, using a caching plugin is s must. Using Cloudflare does not bring much to it (and it can be even worse a bit in terms of TTFB and speed).

GreenGeeks: Detailed results how Cloudflare changes TTFB and speed

Chronological observations

Here are four charts of observations displayed in the chronological order of tests (i.e. no special sorting). Each chart displays a testing configuration (with or without Cloudflare and with or without a caching plugin).

The chart above represents pure hosting performance – no Cloudflare and no caching plugin are used. It looks not as fast as other hosts, but still not bad. There are several occasional peaks (both speed and TTFB). Anyway, I’ll analyze stability and peaks a bit further.

Added now a caching plugin (the chart is above). And the TTFB results look better. We’ll see it in details further.

Removed caching plugin and added Cloudflare (see the above chart). Both TTFB and speed got worse.

Adding a caching plugin additionally to Cloudflare (the chart above) improves the situation compared to the previous chart. Especially TTFB became better.

And now all the chart in one place (just for convenience):

A noticeable difference can be seen in TTFB changes.

Now let’s see basically the same TTFB and website speed data but from a different perspective.

Sorted series

Let’s sort the series by value to see the distribution of values.

On the chart above we see that when a visitor who is located close to the hosting server first opens a small website hosted on GreenGeeks, there’s almost no difference whether the website uses Cloudflare or a caching plugin. The speed is more or less the same at different website configurations.

Time To First Byte at the first visit is very stable and very good if a caching plugin is used. If a caching plugin is not used, TTFB is not that brilliant (mostly TTFB tests are still okay though).

Repeat View has a very similar pattern as the First View. The only significant difference is that repeat view speed is better (for all website configurations).

TTFB repeat view is almost identical TTFB First View results. I.e. a caching plugin improves the results noticeably.

And here’s all four charts in one place for easier comparison:

Sorted series without peak values

For better view on how Cloudflare (and a caching plugin) influence website performance including TTFB, I remove the peak values. Here are the result charts.

As you can see, in general, there’s no big difference in website configuration. But a little bit worse speed values for first view are achieved if using Cloudflare without a caching plugin.

Although speed at different website configurations does not differ very much (see the chart above the last chart), TTFB does differ (see the last chart above).The best TTFB is achieved if using a caching plugin. Cloudflare makes TTFB worse compared to using a caching plugin.

The chart above with peaks excluded demonstrates that tests give very similar results for whatever website configuration. A website without Cloudflare or a caching plugin is just a little bit slower than at other configurations though.

Time To First Byte at repeat view demonstrates the same pattern as for TTFB at the first view. I.e. website configuration with a caching plugin is better.

Analyzing peaks

The speed peaks at GreenGeeks were less than 1% of all tests. The website configurations with Cloudflare positively influenced on website speed at the first view.

Any configurations caused TTFB peaks at the first view. “No Cloudflare with a caching plugin” caused more peaks, but all but one of these peaks are very small (less than 0.4 seconds).

Just a couple of speed peaks at repeat views. Not a big deal.

There are more TTFB peak values at repeat view. And lots of peaks belong to the configurations where neither Cloudflare nor a caching plugin are used. Although most of the peaks are not that large (TTFB less than 0.6 seconds is totally fine).

Here’s a summary chart and table displaying the number of peak occurrences:

As you can see, the configurations with just Cloudflare gives no peaks. And although this configurations gives comparative worse TTFB, it gives a better speed (and no peaks).

The rest of the charts with tables in this section are the derivatives from the above table data for GrenGeeks. Unlike other hosts, this section is not much to analyze.

As you can see, if looking from a general perspective, using no Cloudflare causes comparatively more peaks. As regards using or not using a caching plugin, the difference is not that big.

Intermediate conclusions for GreenGeeks

When testing GreenGeeks’ Time To First Byte and speed at different website configurations (with and without Cloudflare and/or caching plugin) the difference between the configurations was not dramatic from a speed point of view.

At the same time, TTFB was better if a caching plugin was used. There was no significant difference for TTFB whether Cloudflare was used or not.

Thus, if you are using GreenGeeks, then using a caching plugin is recommended to improve TTFB.

Conclusions

I have run several thousands of tests to explore how Cloudflare affects Time To First Byte (TTFB). Also, paying attention to website speed (full page load time). I have taken in to account first and repeat views.

I have tested small websites on three difference hosts (SiteGround, HostWinds and GreenGeeks).And found out that Cloudflare influences TTFB in different ways. The intermediate conclusions for this research for the hosts are here: for SiteGround, for HostWinds, for GreenGeeks.

The main general conclusion is the following. There is a common pattern that Cloudflare increases a number of peaks. In other words, there are few percent of tests which reveal the comparatively large TTFB (and consequently, slower speed) because of using Cloudflare. Two of the three hosts clearly show this dependence.

The second important conclusion is that a caching plugin is definitely a positive factor in website configuration. It allows improving both Full Page Load Time (website speed) and Time To First Byte (TTFB).

In general, SiteGround appeared to be the fastest of the tested hosts. Using Cloudflare made my site a little bit faster in general, but at the same time it caused more peaks. After all, both TTFB and speed of the website hosted with SiteGround were superb.

HostWinds has not bad TTFB in my tests. But TTFB becomes significantly better if a caching plugin is used. Cloudflare does not influence TTFB (as well as speed) to a significant extent. At the same time, using Cloudflare increases a number of peaks.

GreenGeeks in my tests did not have a lot of peaks. Cloudflare did not affect TTFB or speed much. Although a caching plugin does improve TTFB to some extent.

Some practical advice regarding Cloudflare:

• Using a caching plugin (whether you use or you don’t use Cloudflare) is a good idea in most cases. A free plugin can work very well.
• If your visitors are close to your hosting server, then test if a free Cloudflare service brings you any benefit (especially if you don’t use Cache-Everything rule). Sometimes using a caching solution provided by a hosting is all you need.
• A free service like Cloudflare comes at a price (it can cause occasional speed peaks, like for a couple of percent of visits). Not a big deal if all other time it improves your website speed.
• If you are in doubt whether Cloudflare bring you advantage, then test it. Use multiple tests with a free testing tool like webpagetest.org, or buy a monitoring service like I do for the hosts I constantly monitor.

Finally, here’s the conclusion that goes without saying. Having a good hosting does play a big role in website speed. Time To First Byte, if it’s within reasonable limits (much less than a second) is not that important for final website speed (a couple of researches on this topic here and here). Some hosts may provide better TTFB but significantly worse speed for your website.

By the way, I monitor performance of several hosts in 24/7/365 manner (website speed is checked every 20 minutes, TTFB is checked every minute). The results since 2016 are here.

P.S.: In this article I focused mainly on Time To First Byte (TTFB). Besides, as my tests in this research show, in many ways Cloudflare does not improve website speed. And even may decrease the speed. This confirms the same conclusion that I made in my other research (tested on another host). However, using a Cloudflare’s Cache-Everything rule does improve the website speed (see this research).

Also, if you are interested in improving your website speed, here are my articles on this topic. Some of the articles from that list I already mentioned in this post.

You can download a PDF version of this article (1.8 MB):

Share the knowledge...

Facebook

Twitter

Linkedin