Cloudways – Which Server Should You Choose? 2021
Everyone wants faster speeds for their website. Including me.
And Cloudways recently added DigitalOcean Premium Droplets to their platform, so I wanted to figure out how it performed and decide whether to switch or keep using Vultr High Frequency.
So I started to do some speed and load tests on my own.
But then, I got carried away.
Eventually, I ended up testing all the servers including Google and AWS. I did many (probably too many) tests over the course of 30+ days.
What I found was very interesting and I want to share the results with you so you can choose the right server for your needs.
We will be taking a look at 5 main categories:
- Speed & Performance
- Stability & Reliability
- Server Location & Availability
- Pricing & Costs
And also answer some most frequently asked questions such as:
- Should you upgrade from DigitalOcean to DigitalOcean Premium?
- Should you upgrade from Vultr to Vultr HF?
- Which Cloudways server is best for WooCommerce/Dynamic Sites?
- How many websites can you host on each server?
Speed & Performance
To get a comprehensive understanding of how fast and how well each server performs, we are going to do 4 different tests.
#1 Website Speed Test – on Blank WordPress Installation
#2 Website Speed Test – on Homepage (w/ WooCommerce, Theme, Plugins Installed)
#3 FastOrSlow Test – check the Largest Contentful Paint score and other core vital Google metrics
#4 Browse & Checkout Test – see how each server performs in a real-world eCommerce scenario
The Setup Process
Here are the basics:
- Setup Servers On All Infrastructures: Linode, DigitalOcean, DigitalOcean Premium Droplets, Vultr, Vultr High Frequency, AWS, and Google Cloud.
- All servers were the smallest 1GB RAM with the exception of Google at 1.7GB Ram & AWS at 2GB Ram 2vCPU
- Server Location All in the US (West Coast) – And 1 East Coast for Linode (Newark)
First Test: Blank WordPress Installation
After launching each of the servers, I did a speed test using Pingdom on the default blank WordPress installation.
All Server & WordPress were the default settings (straight out of the box).
Unlike other reviewers which just test 1 location – I tested 7 locations, each 3-4 times, and then averaged:
- Tokyo, Japan
- Sydney, Australia
- London, UK
- Washington, USA
- San Francisco
- Saul Paulo, Brazil
Because most website visitors come from all over the world, and this gives you a much more accurate look at the load times.
And I did 3-4 tests because sometimes it can vary during the day and sometimes the first test is not accurate since it may not be cached.
|Server (1GB RAM)||Location||Avg Load Time (secs)|
|Linode 1||Newark,NJ (East Coast USA)||0.65|
|Linode 2||Fremont, CA||0.77|
|Google (1.7GB Ram)||Las Vegas||0.71|
|AWS (2GB RAM, 2vCPU)||California||0.71|
|DigitalOcean Standard||San Francisco||0.72|
|DigitalOcean Premium Droplets||San Francisco||0.73|
|Vultr High Frequency||Silicon Valley||0.77|
Summary & Thoughts
All servers performed very well – with average loading times between 0.70-0.77 secs. To give you context some of the best hosting providers such as SiteGround & Hostinger load at around 1 sec+ for their shared hosting plans.
One standout exception, which was Linode (Newark) loading on average at 0.65 seconds. My initial thoughts were since it’s on the East Coast of the US. The loading times for the UK and Europe are better – therefore lowering the overall average.
However, I set up two additional servers – Vultr HF server in New York and Linode in Fremont, California, and the average loading time was 0.78 and 0.77, respectively.
So, it could be perhaps that specific Linode (Newark) server specifications/resources are different.
Second Test: Basic Woocommerce Site
So, next, I installed a basic eCommerce theme, installed WooCommerce plus a few plugins, and added in a few products. All with the same settings again and used WP Fastest Cache.
So now, I wanted to see how fast it loads a basic page with some content on it.
Again, each page was tested 3-4 times at 7 locations using Pingdom.
Page size was around 690KB and 48 Requests
Here are the results:
|Server (1GB Ram)||Location||Avg Load Time (secs)|
|Linode 1||Newark, NJ (East Coast USA)||1.10|
|Linode 2||Fremont, CA||1.35|
|Google (1.7GB Ram)||Las Vegas||1.22|
|AWS (2GB RAM, 2vCPU)||San Jose, CA||1.20|
|DigitalOcean Standard||San Francisco||1.27|
|DigitalOcean Premium Droplets||San Francisco||1.24|
|Vultr Standard||Los Angeles||1.43|
|Vultr High Frequency||Silicon Valley||1.40|
Summary & Thoughts
- Linode (Newark) performed the best with an average load time of just 1.10 seconds.
- DigitalOcean Standard & Premium Droplets performed better than Vultr and Vultr HF.
- There isn’t much difference between DigitalOcean Standard and Premium. With Premium Droplets just 20ms faster.
- There isn’t much difference between Vultr & Vultf HF either – only 30ms faster.
Third Test: FastorSlow
For this test, I mainly wanted to see the Largest Contentful Paint Scores & other important metrics which were included in the latest Google’s Core Web Vitals update. This website also tests 18 locations around the world – so we get an accurate look at the overall performance since your website visitors don’t just come from one place.
Largest Contentful Paint (LCP) basically means how long it takes the largest piece of content to load. For example a Hero Image or a Large Block of Content. This is important as Google doesn’t want users waiting and just looking at a blank page.
A Good Largest Contentful Paint is 2.5 secs or less.
Here are the results
Linode 1 (Newark) – on top again with a 95 Overall Score, LCP Average secs = 1.36 secs (Low of 498.10ms and High of 2.22secs), Average TTFB = 420.41ms
Linode 2 (Fremont) – Overall Score of 89, LCP Average = 2.01 sec (Low of 810.16ms and High of 3.25secs), Average TTFB = 499.99ms
DigitalOcean Standard – Overall Score of 90, LCP Average = 1.56 sec (Low of 573.47ms and High of 2.41secs), Average TTFB = 565.13ms
DigitalOcean Premium Droplets – Overall Score of 91, LCP Average = 1.56 secs (Low of 545.34ms and High of 2.46secs), Average TTFB = 524.33ms
Vultr Standard – Overall Score of 89, LCP Average = 1.50 secs (Low of 683.91ms and High of 2.22secs), Average TTFB = 603.45ms
Vultr HF – Overall Score of 92, LCP Average = 1.52 secs (Low of 567.08ms and High of 2.33 secs), Average TTFB = 488.74ms
Google Cloud – Overall Score of 91, LCP Average = 1.53 secs (Low of 655.59ms and High of 2.51s), Average TTFB = 474.69ms
AWS – Overall Score of 93, LCP Average = 1.52 secs (Low of 589.74ms and High of 2.51s), Average TTFB = 461.92ms
Tip: Slow LCP times are often caused by unoptimized images or slow hosting providers
Keep Hero Images less than 500KB and use JPEG file format.
Summary & Thoughts
#1 A good LCP score is less than 2.5 secs. So all servers pass that test easily with an average around ~ 1.5 sec+.
#2 Linode (Newark) performed the best with an LCP average of just 1.36secs.
#3 However, Linode (Fremont) performed the worst with an LCP average of 2.01 – so if you choose a Linode server – doesn’t necessarily mean you’ll get super-fast speeds like the one in Newark.
So does it mean that you should select Linode? and DO & DO Premium?
These first three tests are just basic page speed tests on a fairly “static” page. Static meaning, the content doesn’t really change. i.e. Small Business website or a blog like mine.
The opposite of static would be a Dynamic site. i.e. WooCommerce, Membership Platform, Directory, eLearning Course Site.
With a static website, it can be easily cached (i.e. put into temporary storage) and then that content can be easily delivered to the next web visitor a lot faster.
So, If you are creating a small business website or a blog – then yes, Linode (Newark) and DO and Do Premium Droplets can be a good option.
However, when it’s a dynamic website such as a WooCommerce/eCommerce site – some things can’t be cached i.e adding products to cart, cart page & checkout, etc.
Every time a visitor performs an action it requires more processing power to deliver the content/page to them since the data is in the main storage/database.
So, we need to do some further tests to come up with a more accurate conclusion which lead us to our next test.
#4 Browse & Checkout Test – WooCommerce
In our previous tests, it was mostly based on a static website (i.e. blank WP install and the homepage of the eCommerce page).
While these tests are good, it doesn’t really give us a full picture of how well the server performs under different conditions such as a customer browsing your shop, adding products to the cart, viewing the cart, and checking out.
Normally, these pages aren’t cached – therefore it requires more CPU power, faster storage, etc to deliver the content to the user.
So we need to do some additional testing to see how well the server performs for a dynamic site. i.e. WooCommerce, Membership, Forums, Learning management systems (LMS), etc.
- Turned off Varnish (Server-side cache) and also WP Fastest Cache was off too – since I wanted to see the raw performance of the server itself.
- I used the k6 browser recorder (chrome extension) to record a few actions 1) Navigate to Shop Page 2) Add Product to Cart 4) View Cart 5) Fill in Info and Checkout
- Then I used k6.io and set the Load to 20 virtual users for 5 Minutes and tested it on all the servers. This will basically simulate 20 users on the website browsing through the pages as recorded for 5 minutes then report on the response times.
Here are the results:
(Lower response times is better)
#1 AWS – 575ms
#2 Vultr HF – 723ms
#3 Vultr Standard – 1042ms
#4 Digital Ocean Premium – 1252ms
#5 Linode 2 (Newark) – 1382ms
#6 Google Cloud – 1407ms
#7 Linode 2 (Fremont) – 1412ms
#8 DigitalOcean Standard – 2531ms
Summary & Thoughts
#1 – While Linode (Newark) performed really well in the basic website speed tests – it didn’t perform as well in our browsing and checkout (dynamic) tests.
#2 AWS came out on top – which basically means for an eCommerce website, you will have the best response times and offer the best experience for visitors. Note: AWS is using 2GB RAM and also has 2vCPU’s (their smallest plan).
#3 Second was Vultr HF with a response time of 723ms – which is very good since the price is around half of that of AWS. And it’s only using 1GB RAM and 1vCPU.
#4 My recommendation for an WooCommerce/Dynamic site would probably be Vultr HF – it’s more affordable than AWS. My second choice would then be AWS. AWS does have some extra benefits such as the ability to scale up and down in server size and they are the most reliable (up there with Google).
#5 On the kernl blog, they also found AWS to be best with 2.83 checkouts per second and Vultr Standard was second with 2.55 checkouts per second. So we can assume Vultr HF will be probably very close in terms of checkouts per second to AWS or even beat it.
OVERALL SPEED & PERFORMANCE RECOMMENDATIONS
We still have a few rounds of testing to do, but from a speed & performance standpoint – here are my recommendations:
#1 – For a Static type of site (the data doesn’t change very often)
e.g. Small Local Business site, which is mostly text and images. Or a Blog similar to mine.
I would probably go for DigitalOcean or DigitalOcean Premium Droplets. And Linode (Newark) if it’s close to my audience.
#2 – For a Dynamic type of site (WooCommerce, LMS, Forum, Directories, Blogs with lots of comments to load)
I would probably go to Vultr High Frequency because the price to performance ratio is very good. And AWS is my 2nd choice because of cost. (If budget is no issue – choose AWS)
#3 What’s a Good All-Rounder Performer?
I’d say DigitalOcean Premium Droplets – they performed fairly well in the static website speed tests and for the Browse and Checkout, they did okay too.
Not all servers offers the same flexibility, so choosing the right one from the beginning can save you time and money.
Let’s take a closer look
#1 All servers can scale up (in server size) BUT not all can scale down.
Let’s say you expect an increase in visitors to your website for a promotion or it’s Q4 (holiday season) – you can easily scale up your server size to handle the traffic.
But what if the traffic dies down during Q1 or after your promotion ends and you don’t need those resources anymore?
Normally you would just scale down with a few clicks. But you can’t if you are using DigitalOcean, Linode, or Vultr.
You would need to 1) Clone your server back to the smaller server size 2) Update your DNS/IP – which could take anywhere between 15-20 minutes to fully propagate/connect.
This is not hard to do but could be a problem for some websites e.g. you have a very high traffic site and 20 mins downtime will cost you thousands.
If it’s an issue, Choose AWS or Google. Those servers will allow you to scale up and down easily.
#2 The secret benefit of DigitalOcean
If you choose DigitalOcean standard and later decide you want extra performance, you can scale up into a DigitalOcean Premium Droplet plan with just a few clicks.
The same can’t be said about Vultr. You will need to clone your server to migrate over.
Stability & Reliability
To find out how stable a server is – we are going to do a load test with 200 users on the website simultaneously.
For this test, I’ll be using k6.io again – which you also sign up for and do load tests of up to 50 virtual users for Free.
Note: Focus on the stability (Standard Deviation) and the max response compared to average response times.
P95 and P99 values are the 95th and 99th percentile values. So basically if the p99 is 58ms – it means that 99% of results had a response time of 58ms or less.
The closer the number to the average the better.
Let’s take a look at the results:
Linode 1 (Newark)
Standard deviation is 10ms, average response is 25 and max of 292ms.
Linode 2 (Fremont)
Standard deviation is 12ms, average response is 9.24ms and max of 324ms.
Standard Deviation is 1.05ms, an average of 6.33ms, and Max of 25ms.
DigitalOcean Premium Droplets
Standard Deviation is 1.59 ms, average of 6.27ms and Max of 111ms.
Standard Deviation is 2.34 ms, an average of 33ms and Max of 172ms.
Vultr High Frequency
Standard Deviation is 0.73ms, average of 6.11ms and Max of 21ms.
Standard Deviation is 3.14ms, an average of 51ms, and Max of 213ms.
Standard Deviation is 2.14ms, an average of 5.17ms, and a max of 51ms.
Summary & Thoughts
Overall, all the servers at Cloudways perform pretty well in the load tests with standard deviation varying from 0.73 to 12ms
- Linode is the weakest among the servers with a standard deviation of 10ms for Newark and 12ms for Fremont. They also seemed to have a high max response time compared to the average.
- Vultr HF performed the best with only a standard deviation of just 0.73ms – meaning it’s very stable and the load times don’t vary much even in high traffic conditions. What’s even more impressive is that the p99 score is 9ms which means that 99% of response times were 9ms or under. Vultr standard performed decently.
- DigitalOcean servers performed fairly well too. Both with decent p95 and p99 scores.
- If I had a choice, I’d probably go with AWS over Google.
- To give you guys context of the performance, I’ve included the results from basic shared hosting with 20 & 50 virtual users below.
Basic Shared Hosting – 20 Virtual Users
Basic Shared Hosting – 50 Virtual Users
While these results are fairly good for shared hosting, it doesn’t compare with the servers on Cloudways when it comes to speed & performance.
Having decent reliability matters.
When a friend says they will meet you at 7 pm for dinner the next day, but cancels last minute – it’s annoying and it can ruin your night.
And if they do it too many times, you probably won’t ask them to come out anymore. Right?
The same for your website – if it goes down once or twice in a year it’s okay – But any more than that you’d probably want to switch. So I wanted to help you make the right decision from the beginning so you can avoid any inconveniences in the future.
Technically, cloud hosting is supposed to have virtually 100% uptime since your website is supported by several servers. But at the end of the day – nothing is perfect and unfortunately, I’ve had a few instances where my site was down 5-15 minutes at a time.
It’s not bad at all if you are comparing it to shared hosting. But it’s important to know.
Luckily, Cloudways have their own incident reporting – where they list any issues they are having. For example – server maintenance, outages, connectivity issues, or other problems.
Based on the past 12 months, here are servers with the least incidents/problems (ranked in order).
#1 AWS & Google – these rarely had any incidents at all and are the most reliable infrastructures on Cloudways.
#4 Vultr – Yes, unfortunately, Vultr has had its fair share of server maintenance, server degradation, and partial outages. Could be due to the popularity of their Vultr High-Frequency servers and they are the newer company (founded in 2014).
Does it mean you should avoid them?
Not really, they are still great in comparison to shared hosting but when we compare it to the other servers in Cloudways – they are the least “reliable”.
To give context, over the last 6 months for Hoganchua.com the uptime has been 99.99%. (Roughly 36 minutes offline over the course of 6 months – which is still okay for me).
But if you are creating a mission-critical website. Just a few minutes of downtime per month can cost thousands, so if that’s the case – I’d consider AWS or Google since they rarely have any issues at all.
Server Location & Availability
Perhaps you’ve decided on a server based on the results (above) of speed and performance.
However, another very important factor you should consider is the server location and availability.
Generally, you should choose a server that is closest to your audience. This means that there is less physical distance for your website’s data to travel – which means a fast load time and great experience for your users.
Let’s say for example: You’ve decided to go with DigitalOcean Premium, but you are a local business in Melbourne, Australia.
Then I wouldn’t recommend DigitalOcean Premium – because they don’t have a server in Australia (as of this writing). Yes, you could add a CDN but it won’t beat having a server in Australia.
Then, in that case, I’d look for and compare the alternative servers which do have servers in Australia. i.e. Linode (Sydney) or Vultr (Sydney).
How about the availability?
What I mean about the “availability” is that sometimes you may choose a server size of 4GB RAM for a certain location. However, the 8GB RAM server is not available due to the servers being full. This would mean, if you need more resources and want to scale up – you may need to scale directly into the 16GB or above costing you a lot more.
Note: Sometimes it may be available when you select your servers – however, once you’ve launched your server – it may show that the “selected size is not available”. So it’s something to keep in mind.
I found Vultr & Vultr HF to have more availability issues than the other providers.
It’s not a big issue, cause you can clone and move to another – but its something to keep in mind.
Some servers perform better for certain locations
Here is some information that I wouldn’t have known – if I didn’t do extensive testing. This is very interesting.
|Linode 1 (Newark)||Linode 2 (Fremont)||DigitalOcean (San F.)||DO Premium||Vultr (L.A)||Vultr High Freq.||AWS|
|Sao Paulo, Brazil||1.26||2.08||1.99||2.01||2.16||2.17||1.86||1.93|
|Avg Load Times (sec)||1.10||1.43||1.27||1.24||1.43||1.40||1.22||1.20|
As you can see from the results – both DigitalOcean servers (based in the US) performed really well for Australia but for Linode, Vultr and even Google or AWS didn’t perform too well in that region. So if you have a big audience in Australia but your Main Audience is the US/UK – DigitalOcean could be a great choice.
And Linode servers – didn’t perform the best in Japan.
So hopefully with the graph above – you can see which server best fits your scenarios and needs.
Pricing & Costs
Let’s go through the pricing and compare them. You can also view the pricing here in detail.
As you’ve probably noticed – AWS and Google are more expensive. But just how much more do you pay for comparative plans on Linode, DO, and Vultr?
AWS – Starts at ~$34 for 2GB & 2vCPU with 20GB storage. Server pricing is different for location e.g. Sao Paulo cost around $58+/mo for the same specs. Bandwidth is pay as you go, so it really can add up. In addition to their base servers they offer “High Computing Instance” servers starting at ~$158 for 4GB RAM & 2vCPU for even more performance.
Google – Plans also start from ~$34 for 1.7GB RAM & 1vCPU with 20GB storage. Server pricing is different for location e.g. Sao Paulo cost around $50+/mo for the same specs. Similar to AWS, Bandwidth is also pay as you go and they also have “High Computing Instance” servers starting at ~$85/mo for 2GB RAM & 2vCPU.
Linode – Starting at $12/mo for the 1GB RAM server & $24/mo for the 2GB RAM server. AWS & Google is around x1.5 more.
DigitalOcean – Starts at just $10/mo for 1GB Ram server, they are the most affordable. Their 2GB server is $22/mo.
DigitalOcan Premium Droplets – Starts at $12/mo for the 1GB RAM server and $26/mo for the 2GB RAM Server.
Vultr – Starts at $11/mo for the 1GB RAM server and $23/mo for the 2GB RAM Server.
Vultr HF – Starts at $13/mo for the 1GB RAM server and $26/mo for the 2GB Ram Server
Summary & Thoughts
- Pricing of Linode, Vultr, and DigitalOcean is very similar – so if you are choosing one, look at how they performed in the speed, performance, and reliability tests rather than selecting based on cost.
- Linode, Vultr, and Digital Ocean servers are much better value then AWS & Google. They compare well in the speed, performance, and WooCommerce tests. In some cases even beating their more expensive counterparts.
- AWS & Google are roughly x1.5-x2 the price of the other servers when everything is taken into account e.g bandwidth, storage etc. The main benefit of these servers is they are super reliable, with barely any issues. They offer larger servers too. And you have more flexibility to scale up and down. And for some, being associated with their brand may be an added benefit.
Okay, so which one would I pick?
#1 – If budget was no issue – and if I was building a mission-critical site (any type), I would choose AWS over Google and others.
#2 – For a Good All-Rounder – I would choose DigitalOcean Premium Droplets since, for the speed test they were pretty good, WooCommerce checkout test was average, good stability and reliability. They also have better availability than Vultr. Pricing is good too.
#3 – For eCommerce/Dynamic Websites – I would choose Vultr HF. They have more processing power and performed well in the WooCommerce test, and were very stable in the load test. The only letdown is their reliability and availability. (Though for most people it should be fine)
Note: Choose a server location closest to your main audience. So if you’re preferred server provider is not available in a certain location – Look at another provider. Overall, all the servers on Cloudways offer x2-3 better performance than shared hosting!
Frequently Asked Questions
1) Should you upgrade from DO to DO Premium?
From my tests, there isn’t too much of a difference in terms of improvement for a static website that can be well cached. But there should be much more improvement for dynamic sites (i.e. eCommerce) because of the newer Intel processors and the NVMe storage.
However, It’s worth noting that DigitalOcean released two premium droplets – One with second-generation Intel Xeon Scalable processors and another with 2nd Generation AMD EPYC™ processors.
But on the Cloudways platform – we only have access to the Intel processors.
The results from another blog, also show that with sites that can be easily cached – there is little to no difference between them. Intel performed slightly better for “non-cached” sites, but AMD performed a lot better.
So what does this mean?
Upgrade, if you want that extra little performance boost or you are a dynamic type of site (e.g. WooCommerce). But if you are a simple business website, then probably not worth the $2/mo increase. Unless in the future Cloudways start offering DO’s Premium Droplets with the AMD processors.
2) Should you upgrade from Vultr to Vultr High Frequency?
In my basic page speed tests of the WooCommerce Homepage – there was only a decrease of 30ms from Vultr to Vultr High Frequency. So not much in it.
But in my WooCommerce Checkout testing – there was a decrease of around 300ms. That is quite a significant amount.
So ultimately, I would upgrade if I am building a dynamic type of website (i.e. eCommerce, Course Platform, Directories, Forums, etc).
Basic websites (small business website/simple blogs) which can be well cached probably won’t see too much of a performance boost.
3) Which Cloudways server is best for WooCommerce/Dynamic Sites?
From my tests – AWS and Vultr High Frequency performed the best.
AWS is the more expensive option but comes with better reliability, the ability to scale up and down, and slightly better speed the Vultr HF.
Vultr High Frequency is not far behind and is a much more economical option start at just $13/mo. Vultr HF uses blazing fast 3.8GHz Intel Skylake Processors and NVMe SSD Storage – which offer greater Read/Write speed than the traditional SSD drives. Additionally – they do have 17 data centers giving you the complete freedom to develop and deploy on the ones closest to your and your clients for low latency.
But sometimes reliability and server availability are not as good as the others.
4) How many websites can you host on a server?
For the 1GB RAM servers, they have around 25-30GB of storage. Generally should be enough for around 5-10 Small websites! So if you are getting the DigitalOcean plan at $10/mo – it’s only $1/mo per website.
Note: Email hosting is not included and can be added and set up within Cloudways at $1/mo per inbox.
Even though you can have “unlimited” websites, once you hit a certain amount of page views – your performance will start degrading and you’ll need to scale up.
How to Get Started on Cloudways
Step 1: Go to Cloudways.com & click on “Get Started for Free” – You get a 3-day free trial (no cc required)
Step 2: Enter your details and Promo Code: HOGAN for 20% off for 3 months. Then click on “Start Free” to sign up.
Step 3: Activate your account (check email) to launch servers.
Step 4: To launch your servers, select app e.g WordPress, enter app and server names. Then select your provider of choice and size of the server.
Note: If you are starting out, most of the time 1GB RAM may be sufficient as it can handle up to 70k page views per month. You can always scale up when needed.
Once launched you can follow the video below – it gives you a better step-by-step process on how to set up your website, migrating your existing site, add a FREE SSL certificate, set up and emails, etc.
TIP: Check the YouTube description for timestamps to skip to exactly what you need.