Cloudways is a managed panel on a virtual machine you rent from someone else. The panel, the backup button, and the support queue are what the invoice buys. The machine under them is a commodity shape.
This review says who that split fits, and who it does not. It is not a signup walkthrough.
Tianwen Network moved 49 sites off Cloudways onto a self-managed VPS. For the same specification, the cost fell by more than half. The condition was a person watching the machines full time.
That move is the evidence. The conclusion comes after it, not before.
This website also publishes a Cloudways registration walkthrough. The page is in Chinese. It contains affiliate links.
The walkthrough and this review disagree on purpose. One shows a signup path. This one says we left, and why the bill changed.
What the affiliate page is, and what it is not
The Cloudways registration walkthrough sits on this domain and contains affiliate links. Clicking through can pay this site a commission. That is a commercial relationship. A review that hides it is advertising with a different headline.
Read that page if you want the clicks of a signup. Don’t read it as a measured comparison. This review does not repeat promotional lines from it. The only hosting figures used here are the 49-site move, and a separate set of machine measurements that were not taken as a Cloudways benchmark.
A search for cloudways web hosting usually means the buyer is comparing that panel with a raw VPS invoice. The spaced query cloud ways hosting is the same product. People split the brand when they type. Both queries land on a managed layer, not on a magic box.
Henan Tianwen Network Technology Co., Ltd. was founded in 2017. Liu Yu is the technical lead.
The company maintains 30 to 40 export sites on an ongoing basis. That count is not the 49. One figure is a migration batch.
The other is how many export sites are in care now. They are separate records. This page does not invent a story that ties them into one client list.
The ledger from the move
Forty-nine sites left Cloudways as a batch. The destination was a VPS we operate ourselves. The specification was the same class of machine.
The invoice dropped by more than half. Nothing in that sentence is a performance trophy. It is a cost line, with a staffing condition attached.
The condition is the whole trade. Someone has to watch the machines full time. Patches, disk, PHP workers, a certificate, a backup that actually restores.
If that person does not exist, the cheaper invoice is a discount you pay back as downtime. We had the person. That is why the move was rational for us.
It is not a universal instruction.
Could a smaller fleet get the same ratio? This record does not say. It says what happened at 49 sites, same specification, with a full-time operator.
Extrapolating the ratio onto one brochure site is how a review becomes a slogan. Don’t.
We still host the affiliate walkthrough after leaving. Both facts stand. Leaving did not obligate us to pretend the panel is useless.
Keeping the link does not obligate us to pretend we still run the fleet there. A buyer can hold both sentences at once.
Who should stay
Stay if nobody on the payroll can be the full-time operator. A one-site or two-site company that sells through the site, and has no one to read a PHP log at night, is buying the queue and the panel. The premium over a raw VPS is the wage you didn’t hire. That can be the cheaper choice even when the invoice looks higher.
Stay if you want a backup button and a ticket form more than you want the nginx file. Some fixes live in that file. If you will never open it, a host that hides it is not blocking you. It is matching you.
Stay if the site is not your operations. A brochure with a handful of URLs can live on a managed panel for years. The failure mode we care about, a catalogue that times out while the office sleeps, may never show up.
Paying for a panel you barely touch is dull. Dull is allowed.
Ask one question before you stay. Who gets the ticket when the site returns a blank page on a Sunday? If the answer is “Cloudways support”, and you are comfortable with their queue, the product matches the sentence. If the answer is a named person on your side who already watches servers, you are paying twice for the same hour.
Who should not
Don’t pick it because a tutorial said the setup was easy. Easy signup is a different property from a rational three-year cost. We used the platform, then moved 49 sites, and the same-spec bill fell by more than half once a full-time operator was the premise.
If you already employ that operator, run the same comparison on your own invoice. Don’t borrow our ratio as your forecast. Borrow the method.
Same specification, both bills, and an honest line for the person’s time.
Don’t pick it because a server city sounds closer to your buyers. Google documents server location, taken from the server’s IP address, as a signal about the intended audience. The same document says some sites use a CDN, or host where the web infrastructure is better, so the location is not a definitive signal.
Country-code domains, hreflang, and the content itself do the explicit work. The page is Google’s note on multi-regional sites. A datacentre dropdown is not a ranking plan.
Don’t pick it to fix a fat template. A product template we inspected carried a 15 MB image in the first screen. Hundreds of product URLs then failed Core Web Vitals.
The server was not the broken part. Changing host, managed or not, does not shrink that file. The published thresholds are an LCP within 2.5 seconds, an INP of 200 milliseconds or less, and a CLS of 0.1 or less.
Four numbers worth asking any host
These four were measured on machines we operate. They are not a Cloudways scorecard. They are the questions a panel has to answer if you want the bill to mean something. If a vendor cannot show them, you are buying a label.
Installed RAM against the live PHP memory_limit. One recorded case is a 1-core, 1 GB machine running three WordPress sites. It stayed unstable.
The site count was the wrong suspect. memory_limit had been set to 1024M, which is the whole box before the operating system and MySQL. A single PHP process used 80 to 124 MB.
Six workers at the top of that range are 744 MB, and the database is not in the sum. The stable change was memory_limit at 256M, and workers cut from 6 to 4.
Does the panel show the live value, or a marketing line called “PHP memory”? Those are different sentences. Ask for the ini value a request actually sees.
Then multiply worker cap by measured process size, not by the ceiling. Compare the product with installed RAM. Leave room for MySQL.
If the product is larger than the box, the host is oversubscribed on paper.
Disk, and what kind of disk. On a multi-site server the disk moved from 91 percent full to 37 percent full. The space did not come back from deleting a website.
A filter plugin had written its cache into the database. One row was 30 MB. A panel that only shows “disk used” hides a table problem.
Before any move, ask for the ten largest tables.
Time to first byte, from outside the machine. On waimaodulizhan.com the homepage TTFB went from 2.14 seconds to 0.067 seconds. The machine stayed.
The change was an nginx fastcgi_cache block. Another site on that machine already had the block. If your panel cannot add that stanza, and will not add it for you, the host is the constraint.
The figure is one homepage. It is not a promise for every URL, and it was not recorded as a before-and-after against Cloudways.
What a move actually tests
Copying files is the easy half. Mail, redirects, and cache fail quietly after a host calls the job done. The longer list of those quiet failures is moving WordPress to a new host.
Read it before you treat a successful file copy as a finished migration. We are not repeating that list here.
Backups deserve a restore drill, not a green icon. A backup that includes 30 MB cache rows will restore the mess and then bill you for disk. A backup that never gets restored is a file, not a plan.
Ask when the last restore was tried, and how long it took. An answer that names a date is worth more than a feature tick.
Ownership belongs in the same conversation. Whose name is on the server account, the domain, and the billing card? A managed panel can make that feel abstract until you need to leave.
The 49-site move was possible because we could leave. If the contract makes leaving a project of its own, price that project on day one.
The capacity questions for an export catalogue, RAM, workers, and TTFB, are written up as WordPress hosting for an export catalogue. Use that page when the site is a product URL farm. Use this page when the decision is Cloudways specifically.
What “watching the machine” means on a Tuesday
The full-time condition is easy to nod at and hard to staff. It is not a person who glances at a status page. It is the person who can change a number and explain the number.
On the 1 GB box, the fix was memory_limit from 1024M to 256M, and PHP workers from 6 to 4. A support macro that restarts PHP does not make that change. A person who can read the ini file does.
The same person looks at table size, not just the disk percentage. A 30 MB row from a filter plugin’s cache is invisible on a pretty graph and obvious in a table list. Leaving it in place means the next backup is heavier, the next restore is slower, and the next host inherits the mess.
Managed or not, someone has to decide the row is garbage. The panel will not feel embarrassed for you.
Cache is the third Tuesday job. The homepage that moved from 2.14 seconds to 0.067 seconds did not get a new CPU. It got a fastcgi_cache block that a neighbouring site already had.
If your operator can add that block, the managed panel’s convenience is competing with a one-stanza fix. If your operator cannot, stop comparing invoices and admit you are buying their operator.
None of those Tuesdays were logged as a Cloudways incident report. Don’t paste them into a chat with their support and expect a matching ticket. They are the shape of work that remains after you leave a panel, and the shape of work you are paying the panel to absorb if you stay. Match the shape to the person you actually have.
A smaller site is a different sum
One WordPress site on a managed panel can be a boring, correct choice. The 49-site figure does not scale down by dividing the savings by 49 and handing you a per-site discount. Shared costs, the operator’s salary, monitoring, and the time to keep PHP and the database boring, don’t divide that cleanly.
A company with a single catalogue should price one machine plus a fraction of a person, then compare that with one panel invoice. If the fraction of a person is a fiction, the panel wins.
Three sites on 1 GB of RAM can be calm once memory_limit matches the box. The same three sites with a 1024M ceiling were not. That is a configuration fact, not a brand fact.
A managed host that sets the ceiling sensibly has already done the job. A managed host that lets you, or a plugin, set 1024M on a 1 GB plan has handed you the foot-gun and a ticket form. Ask which of those you are buying.
Disk alerts deserve the same split. Ninety-one percent full looks like “buy a bigger plan”. Thirty-seven percent after deleting cache rows looks like “stop storing scratch data in MySQL”.
A support agent who only upgrades the disk will spend your money. An operator who lists the largest tables might delete a plugin’s scratch pad. You want to know which behaviour the plan defaults to.
What this review will not say
It will not say Cloudways is slow, or fast. We do not have a published measurement that isolates their network. Inventing one would make the 49-site ledger look like a trick.
It will not quote a plan price. Prices move, and a stale number helps a sales call more than a buyer. Compare your current invoice with a same-spec alternative, and add the operator’s time. That arithmetic ages better than a screenshot of a pricing page.
It will not promise a ranking, a traffic line, or an enquiry count. Google’s hiring guidance says no one can guarantee a number 1 ranking, and tells you to beware of a guarantee, a claimed special relationship, or a priority submit. We are not going to launder that warning into a hosting pitch.
It will not tell you to hire us because we left. The fit test is the operator. If you have one full time, price a VPS against the panel on the same specification.
If you don’t, the panel is the product, and the affiliate walkthrough is there, with the commercial relationship printed on this page in plain words. Contains affiliate links. We moved 49 sites anyway.
Questions before you stay on Cloudways
Does this Cloudways review contain affiliate links?
Yes. The registration walkthrough on this website contains affiliate links. Clicking through can pay this site a commission. The review says that in the body because the commercial relationship is part of the decision, not a footnote.
Why did Tianwen Network leave Cloudways?
Tianwen Network moved 49 sites from Cloudways to a self-managed VPS. For the same specification, the cost fell by more than half. The condition was a person watching the machines full time. Without that person, the lower invoice is not the same offer.
Does the server's country decide rankings?
Google says server location, taken from the IP address, can be a signal about the intended audience. The same document says a CDN, or hosting chosen for better infrastructure, means the location is not a definitive signal. A datacentre dropdown is not a ranking plan.
Will changing host fix a 15 MB image?
No. One product template carried a 15 MB image in the first screen. Hundreds of product URLs then failed Core Web Vitals. The server was fine. The published targets are an LCP within 2.5 seconds, an INP of 200 milliseconds or less, and a CLS of 0.1 or less.