Cloud server or VPS
A cloud server and a VPS are usually the same KVM virtual machine. The difference lies in what surrounds it: a cloud sells resources by the hour through an API and offers network disks, snapshots and ready-made services, while a classic VPS is a fixed monthly plan with a local disk and a traffic allowance. Here is which of this you actually need and what it costs.
Hosts put the word "cloud" on all kinds of plans, so the name alone says little. It is more useful to look at how the server is billed, where its disk lives and what you can do without contacting support. Below is how a cloud platform differs from a classic VPS, what equal configurations cost according to the accsly.io catalog and where the extra costs of a cloud hide.
What a cloud server is
A cloud server (an instance) is a virtual machine that you create, change and delete yourself, in a panel or through an API. Inside it is the same full virtualization as a good VPS; what differs is everything around it:
- Paying for time. Resources are billed by the hour, and on some platforms by the second: for example, at one of the largest international platforms Linux machines are billed per second with a one-minute minimum. Delete the machine and the charges stop.
- API and automation. Machines, disks, networks and addresses are created with API requests or from Terraform. This matters when servers are spun up for a job and disappear after it: builds, tests, data processing.
- Resizing. Processor and memory change with a reboot, and a disk grows without reinstalling.
- Snapshots and images. A disk snapshot takes minutes, and from it you can bring up a copy of the server or roll back after a failed update.
- Ready-made services. Object storage, managed databases, load balancers, private networks between machines.
A classic VPS is a plan with fixed resources and a monthly price. The disk is usually local, NVMe or SSD inside the same server, traffic comes as an allowance or unmetered, and management goes through the host's panel. Many hosts have added hourly billing, snapshots and an API to their VPS plans, so the line between the two models is blurred. How VPS and VDS are organised is covered in the article VPS or VDS.
Network disk and local disk
The main technical difference of a cloud is where the disk is stored. In a cloud it is usually a volume in distributed storage. For example, Ceph RBD block devices, according to the project documentation, take up space as data is written, spread data across many storage nodes and support snapshots and replication. Hence the cloud's advantages: the disk survives the failure of a physical server, the machine can be restarted on another host, and a snapshot does not require a shutdown.
The price of this is latency. Every write travels over the network to several nodes, so the same database query runs slower on a network disk than on local NVMe. For a site or an API this is usually unnoticeable; for a busy database it is noticeable. Some platforms offer fast local disks, but without restarting on another host. How to test a disk before paying is explained in the article How to check VPS ping and speed, and the difference between NVMe and SSD in NVMe or SSD.
Object storage and managed databases
Object storage holds not a disk but files, called objects, in containers called buckets, and you work with it over HTTP. The most common interface is S3: according to the S3 documentation, an object is a file with its metadata, a bucket is a container for objects, and access goes through a REST API. S3-compatible storage is offered by many cloud platforms and by some VPS hosts. It is a convenient place for backups, images and user uploads, so the server's disk does not keep growing. A backup scheme is described in the article VPS backups.
A managed database is PostgreSQL, MySQL or Redis whose updates, replicas and backups are handled by the platform. You pay more than for the same database on your own VPS, but you do not administer it. If you install the database yourself, see the selections VPS for a database and VPS for PostgreSQL.
What an equal configuration costs
The accsly.io catalog has no separate "cloud" label, so we compared two groups of hosts as of November 2026. Cloud-type: 94 hosts that have both hourly billing and an API. Classic: 152 hosts that have no hourly billing. We took VPS plans in stock with monthly billing. The catalog keeps prices in roubles; the dollar figures are converted at the Central Bank of Russia rate of 4 November 2026 and rounded. Prices are per month; whether VAT is included depends on the host.
| Configuration | Median plan: cloud-type | Median plan: classic | Cheapest plan per host, median: cloud-type / classic |
|---|---|---|---|
| 2 vCPU, 4 GB | $29 (185 plans) | $19 (175) | $15.50 / $15 |
| 4 vCPU, 8 GB | $63 (192) | $38 (167) | $36 / $33 |
| 8 vCPU, 16 GB | $170 (135) | $68 (110) | $75 / $62 |
If you compare the hosts' cheapest offers, the difference is small, from 2 to 22%. But a typical cloud-type plan costs one and a half to two and a half times more: these hosts' line-ups have more expensive variants of the same configuration, and those are what pull the median up. So compare specific plans, not the word "cloud". Classic VPS plans have another saving too: according to the catalog, for 23% of monthly-billed plans a month comes out cheaper if you pay six months or a year in advance. When hourly billing beats monthly billing is covered in the article Hourly billing: when it pays off.
Outbound traffic
With a classic VPS, traffic is usually included in the price. According to the accsly.io catalog for November 2026, of the 20 379 VPS plans that state their traffic terms, 8 762 have unmetered traffic, and the rest have a median allowance of 5 TB a month. Cloud-type hosts in the catalog follow a similar model: a median allowance of 6 TB, with unmetered traffic on about a third of plans.
Large cloud platforms work differently: only a small share of outbound traffic is free, and every gigabyte beyond it is paid for. Both at the largest international platform and at a large cloud platform in Russia, the first 100 GB of outbound traffic a month are free. The Russian platform also does not charge for inbound traffic or for exchange between its services over internal addresses. For a site with moderate traffic this is not much, but for serving files, video or backups to external storage, the traffic bill can exceed the price of the machine itself. How to estimate your volume is explained in the article Bandwidth and traffic: what 1 Gbit/s and unmetered mean.
What else a cloud charges for
- A public IP address. Clouds often bill it separately. In the documentation of the same Russian platform, the address of a running machine costs about $0.0032 an hour including VAT, around $2.40 for 730 hours, while a reserved but unused address costs about $0.0074 an hour.
- Disks and snapshots. They are billed by size and time, separately from the machine. Under the same platform's rules, the time a machine is stopped is not billed, but disks are charged whether the machine is running or not, and together with snapshots they keep being charged even after the machine is deleted. What is charged for a stopped server is covered in more detail in the article on hourly billing.
- Services around the machine. A load balancer, NAT, a managed database, object storage, each with its own price list.
So the cost of a cloud is counted across all the lines, not by the machine price in the calculator. With a classic VPS the total is usually just the plan price, plus extra addresses and backups if they are billed separately; that is the subject of the article How to read a host's plans.
Which to choose when
A cloud server fits if:
- you need servers for hours or days, for tests, builds or one-off calculations: selections VPS with hourly billing and VPS for CI;
- your infrastructure is described in code and created through an API or Terraform: the selection VPS with an API;
- the load changes and you need to add machines quickly, for example for Kubernetes: VPS for Kubernetes;
- the disk must survive a host failure, and you need snapshots without a shutdown.
A classic VPS fits if:
- the server runs all year with a steady load: a site, a bot, a database, mail;
- you need a predictable bill and plenty of traffic: the selection VPS with unmetered traffic;
- the database is sensitive to latency and benefits from local NVMe: VPS on NVMe;
- the budget matters more than automation: for the same money a classic plan usually gives more resources.
A common compromise is a classic VPS for the permanent workload plus object storage for backups and files. And if a virtual machine is not enough, compare it with a physical one in the article Dedicated server or VPS.