Question 1

What is a VPS, and how is it different from shared hosting?

Accepted Answer

A virtual private server (VPS) is a virtual machine sold with dedicated allocations of vCPU time, memory and storage on a shared physical host. You control the operating system, install your own software and manage the network stack. Shared hosting puts many sites on one managed server and hides the operating system from you. A VPS costs more and requires administration, but it gives predictable resource limits, root access and the ability to run containers, daemons and custom runtimes.

Question 2

Do I need a dedicated server instead of a VPS?

Accepted Answer

Choose a dedicated server when you need all of a machine's physical cores and memory, predictable single-tenant performance, or hardware features a virtual machine cannot expose. Choose a VPS when your workload fits within the plan's resources and you value fast provisioning and lower cost.

Most small websites, APIs, containers and staging environments do not need dedicated hardware. If you are considering it only because of a noisy-neighbour suspicion, measure CPU steal and disk latency on the VPS first.

Scope: Sizing guidance, not a performance measurement.

Limits and exceptions

We do not publish independent performance benchmarks for either option.

Related: “vCPU” · “RAM” · CPU-oriented workload · Small database

Updated: 2026-10-25

Question 3

What does a VPS plan usually include, and what costs extra?

Accepted Answer

A typical plan includes a vCPU allocation, memory, a system disk, at least one IPv4 address, some transfer allowance and access to a control panel or API. Provider documentation and this catalog record each of those separately because they are not universal. Common extras are additional IPv4 addresses, block storage, backup or snapshot add-ons, control-panel licences, more transfer and premium support. A price comparison that ignores those lines is incomplete.

Question 4

When is a VPS the wrong choice?

Accepted Answer

A VPS is a poor fit when you need a managed application platform, elastic autoscaling without operational work, or strict compliance controls that require a dedicated environment. It is also wrong when nobody will patch the operating system or watch backups. If your service must stay online through a single-machine failure, a single VPS is the wrong architecture regardless of the plan's quality; you need redundancy at the application or provider level.

Question 5

Should I pay monthly or annually?

Accepted Answer

Annual payment usually lowers the effective monthly price, but it locks money and trust into one provider for a year. Monthly billing costs more per month and preserves the option to move if the service or the price changes. Compare the annual effective rate against the monthly rate and ask what happens on renewal. A discount that only applies to the first term is a promotion, not a long-term price.

Question 6

How should I calculate the real 12-month cost of a plan?

Accepted Answer

Total the promotional months at the promotional price, the remaining months at the renewal price, add any setup fee and any mandatory add-ons such as a primary IPv4 address, then compare that number against your ceiling. A month-to-month plan is simply twelve times its monthly price. The cost calculator in this site performs this calculation from the recorded price fields and shows both the cash total and a planning figure that takes the higher of the effective and renewal monthly price.

Question 7

How are EUR and USD prices compared without misleading me?

Accepted Answer

This dataset records each offer in its native currency and normalizes EUR prices to USD with a dated ECB reference rate. The conversion is for filtering and comparison, not a promise of the amount your card will be charged. Card issuers add their own conversion spread, and some providers charge in a currency different from the one displayed. Always check the checkout currency and the amount actually billed.

Question 8

Which fees are commonly missing from a headline price?

Accepted Answer

Frequent omissions are a primary IPv4 address billed separately, block storage for providers that do not bundle a disk, backup or snapshot add-ons, setup fees, VAT or sales tax, payment-method surcharges and overage charges for transfer. This catalog records setup fee, tax scope and address inclusion when the provider states them, and leaves them unknown otherwise rather than assuming zero.

Question 9

Why is the renewal price more important than the promotional price?

Accepted Answer

A promotional price applies for a limited term; the renewal price is what you pay afterwards. If the promotion covers three months of a twelve-month horizon, nine months are billed at the renewal rate, which can change the ranking of plans completely. The recommendation engine and the cost calculator both use a renewal-aware basis so a teaser rate cannot win on price alone.

Question 10

Is a first-term discount the same as a lower regular price?

Accepted Answer

No. A first-term discount changes only the first invoice; the regular price returns on renewal. A lower regular price means the standard rate itself is lower. The two look similar in a headline and behave differently over a year. Record both numbers before deciding, and check whether the discount depends on a minimum term such as twelve or twenty-four months.

Question 11

What are the risks of prepaying for a year or more?

Accepted Answer

You take provider-default risk for the whole term: if the service degrades, the company changes ownership or you need to move, the prepaid balance may not come back. You also lose flexibility if your requirements change. Prepay only an amount you could lose without breaking the project, and keep an off-provider backup so a billing dispute cannot take your data with it.

Question 12

How do I notice that a provider changed its price?

Accepted Answer

Keep the verification date attached to the price you relied on and re-open the provider's own pricing page or API before a purchase or renewal. This site shows the observed date on every plan and warns when snapshots differ. If a price changes, the older observation is not wrong; it is historical. The correct action is to record a new snapshot, not to keep quoting the old figure as current.

Question 13

What does vCPU actually mean on a VPS?

Accepted Answer

A vCPU is a scheduling unit presented to the virtual machine. On shared plans it represents a slice of physical core time, not a guaranteed physical core; on dedicated plans the provider promises a core or thread that is not shared with other customers. Two plans advertising two vCPUs can therefore deliver different sustained performance, which is why plan family and provider wording matter as much as the number.

Question 14

Does a higher CPU clock speed mean a faster VPS?

Accepted Answer

Not by itself. A higher sustained clock can make single-threaded work faster, but a shared plan can burst above its baseline briefly and then fall back, and memory or disk limits often dominate real workloads. Compare the plan family, the sustained allocation and the storage type before the clock number, and measure your own workload rather than relying on a spec sheet.

Question 15

How can I tell whether a plan uses shared or dedicated CPU?

Accepted Answer

Check the provider's plan family name and documentation: lines described as shared, burstable or regular performance are shared; lines described as dedicated or dedicated CPU are not. This catalog records the family and uses it in recommendations when the provider states it. If the provider does not state the CPU model, the field stays unknown; do not assume dedicated from a high vCPU count.

Question 16

Does it matter whether a VPS is KVM or container-based?

Accepted Answer

Yes, if you need custom kernels, VPN protocols, nested virtualization or full control of networking. KVM gives each instance its own kernel; container-based virtualization shares the host kernel and constrains those use cases. Where a provider does not document the virtualization type, this catalog marks it unknown instead of guessing.

Question 17

When does an arm64 VPS make sense?

Accepted Answer

arm64 makes sense when your software builds or ships for it and your container images publish arm64 variants. It often provides more cores per euro for compiling and modern language runtimes. It is a poor choice when you depend on x86-only binaries, proprietary agents or tooling that assumes x86. Check the architecture field and test the exact image before migrating.

Question 18

How much RAM does a small website actually need?

Accepted Answer

A static or lightly used CMS site can run in one gigabyte of RAM, with enough headroom for the web server, PHP or Node process and the database. Development environments, container hosts and API backends usually need at least two gigabytes because runtimes and build steps compete for memory. Size for the next six to twelve months and watch memory pressure after launch; upgrading later is usually cheaper than overpaying from day one.

Question 19

Why do some pages say GB and others GiB?

Accepted Answer

GB is a decimal unit (one billion bytes) and GiB is binary (1,073,741,824 bytes). Providers mix the two in marketing copy, so a "1 GB" plan may actually deliver slightly less or more than a strictly decimal gigabyte.

This catalog records the provider-stated figure in megabytes and does not silently convert between the two conventions.

Scope: Unit interpretation across provider pages.

Limits and exceptions

The difference is small for small plans but grows with disk sizes.

Related: “RAM” · “NVMe”

Updated: 2026-10-25

Question 20

Is NVMe always faster than SSD in practice?

Accepted Answer

NVMe is a faster protocol than SATA, but the plan's disk size says nothing about the IOPS or throughput limits the provider enforces. Two NVMe plans can perform very differently under database writes. Treat the storage type as a hint, not a benchmark, and measure disk latency on the plan you intend to use.

Question 21

What does "block storage billed separately" mean for my budget?

Accepted Answer

It means the instance price excludes persistent disk. You pay for the compute plan and then for the volume attached to it, often per gigabyte per month. A plan that looks extremely cheap because it has no bundled disk is not comparable to a plan with storage included. Check whether the recorded price includes a usable disk before ranking it.

Question 22

What is the difference between traffic allowance and bandwidth?

Accepted Answer

Traffic is the total data volume allowed per month, usually counted outbound. Bandwidth or port speed is the maximum rate at which data can move at any moment. A plan can have a fast port and a small monthly allowance, or a slow port with no metered cap. Both numbers matter, and this catalog records them separately.

Question 23

Is "unlimited traffic" really unlimited?

Accepted Answer

Usually not. Providers advertising unlimited traffic typically apply fair-use policies that restrict sustained heavy egress, and some charge for bandwidth above a hidden threshold. The plan may also cap the port speed instead of the volume. This catalog records the allowance as unknown rather than infinite when the provider only states fair use, and the recommendation engine does not treat unknown as unlimited.

Question 24

How much does port speed affect a small website?

Accepted Answer

For a small site, request latency and concurrency matter more than raw port speed; most pages are far below even a 100 Mbit/s ceiling. Port speed becomes decisive for large downloads, backups and replication. If your workload moves large files, compare the stated port speed and any transfer cap together.

Question 25

Can egress cost more than the server itself?

Accepted Answer

Yes, for providers that charge per gigabyte beyond the included allowance, especially for media or backup traffic. A cheap instance with expensive egress can cost more than a larger instance with generous transfer. Estimate monthly outbound volume before choosing, and check whether the allowance is per region or per account.

Question 26

What is the difference between a region and a datacenter?

Accepted Answer

A region is a geographic grouping used for latency and compliance; a datacenter is a specific facility inside it. Providers often list regions in marketing and datacenters in APIs or order forms.

This catalog records the provider's datacenter label and maps it to a region for filtering. A provider region catalog does not guarantee stock for a given plan at a specific facility.

Scope: Location data in this dataset.

Limits and exceptions

Exact per-size stock is rarely published.

Related: Guides: Regions, datacenters and latency · “Unknown (data gap)”

Updated: 2026-10-25

Question 27

How do I choose a region for latency?

Accepted Answer

Start from where your users are, then test: provision a small instance in the candidate region and measure round-trip time from representative locations. Published region lists tell you where infrastructure exists, not what latency you will see. If your audience is split, consider two small instances rather than one remote one; the second instance often costs less than the latency penalty.

Question 28

Can a plan be unavailable in a listed region?

Accepted Answer

Yes. Region catalogs list where a provider operates, not which sizes are in stock at each facility. Some providers also price regions differently or restrict specific plans to core regions.

Verify availability for the exact size and region at checkout; this catalog records the evidence type for each location so you can see whether it came from exact datacenter pricing or a region catalog.

Scope: All offers with location data.

Limits and exceptions

Stock changes continuously and is not tracked in real time.

Related: “Unknown (data gap)”

Updated: 2026-10-25

Question 29

Should I deploy in more than one region?

Accepted Answer

Deploy multi-region when latency or availability requirements justify the operational cost of replication, DNS failover and data consistency. For a small site with one audience, a single well-placed instance is usually enough. Multi-region doubles the number of failure modes; plan the data layer before adding instances.

Question 30

Is an IPv4 address always included in the plan price?

Accepted Answer

No. Some providers include a primary IPv4 address, some bill it as an add-on and some quote the server price without it. Hetzner's price API, for example, quotes the server without the primary address. Check the plan's recorded price composition and the provider's checkout before comparing totals.

Question 31

What IPv6 allocation should I expect?

Accepted Answer

Allocations range from a single /128 address to a routed /64 or /48 prefix. A /128 is enough for one service; a /64 or /48 lets you address containers or downstream networks. This catalog records the allocation only when the provider documents it, and the IPv6 filter distinguishes unknown data from confirmed support.

Question 32

How do I verify that a VPS is really dual-stack?

Accepted Answer

Provision the instance, then test inbound and outbound connectivity for both address families from an external network. A provider stating "IPv6 supported" does not guarantee an address is assigned by default. Check reverse DNS if you run mail, and confirm the prefix is routed rather than link-local before migrating a production name.

Question 33

Why does reverse DNS matter for a VPS?

Accepted Answer

Mail servers increasingly reject connections from addresses without matching reverse DNS. Some providers let you set rDNS in the panel or API, and some only do it on request. Confirm the provider supports rDNS for the plan and address you buy; a dual-stack address does not automatically come with PTR records for every address in the prefix.

Question 34

Is a provider snapshot the same as a backup?

Accepted Answer

A snapshot captures a disk state and is usually restored by creating a new instance from it. A backup is a recovery strategy: it includes retention, off-provider copies and a tested restore procedure. A snapshot stored only inside the provider account disappears with a billing or account problem. Treat it as one layer, not the whole plan.

Question 35

Why does retention matter more than backup frequency?

Accepted Answer

Frequency determines how much recent work you can recover; retention determines how far back you can go. A daily backup kept for seven days cannot recover a file deleted nine days ago. Decide the maximum acceptable data loss and the longest history you may need, then confirm the provider's schedule and retention can meet both.

Question 36

Why keep a backup outside the provider?

Accepted Answer

An off-provider copy survives account suspension, billing disputes, region-wide failures and provider changes. It is also the only copy you can restore without the provider's cooperation. Encrypt downloaded archives and keep the key somewhere that is not the server being backed up.

Question 37

How often should I test a restore?

Accepted Answer

Test after any change to the backup configuration, and at least quarterly for data you cannot rebuild. A restore test should prove that a known file, a database dump and a full instance can be recovered within your targets. Record what was restored, how long it took and what failed. An untested backup is a hypothesis.

Question 38

Can I get a refund if I cancel?

Accepted Answer

Refund terms are provider-specific and often differ by product, billing term and payment method. A money-back window may apply to one plan family but not to prepaid annual contracts or crypto payments.

This catalog records refund terms only where they were captured; for the rest, treat them as unverified and read the provider's terms before paying.

Scope: Provider-specific terms.

Limits and exceptions

Refund policies are not captured for every provider in this dataset.

Related: “Unknown (data gap)”

Updated: 2026-10-25

Question 39

What does an SLA actually guarantee?

Accepted Answer

An SLA typically states an availability target and a credit mechanism, not a refund. Credits are often small, require a claim within a window, and exclude scheduled maintenance and customer-caused outages. Read the definition of downtime and the claim process before relying on the SLA for a business-critical service.

Question 40

Which terms-of-service rules commonly affect VPS use?

Accepted Answer

Common constraints include prohibited content, resource-abuse rules, outbound mail restrictions, port scanning limits and acceptable-use requirements for crypto mining. Enforcement can suspend an instance without notice. If your workload touches these areas, read the provider's terms before deploying; a lower price does not offset a suspension.

Question 41

What happens to my data when I cancel?

Accepted Answer

Most providers reclaim the instance and its disks shortly after cancellation, and snapshots may be deleted with it. Some keep a short grace period; others do not. Export the data and confirm the cancellation email before the billing term ends. Do not rely on a provider's retention after cancellation as your backup.

Question 42

How should I size a VPS for a small business website?

Accepted Answer

Start with the CMS or framework's minimum requirements, add the database, then size for expected concurrent users and page weight. A small CMS with caching typically fits in one to two gigabytes of RAM and one shared vCPU. Plan for peak traffic and backups rather than average load, and measure after launch before upgrading.

Question 43

What does a small database need from a VPS?

Accepted Answer

The working set should fit in RAM, storage should be fast and reliable, and backups must be consistent. For a small production database, start at four gigabytes of RAM with two vCPUs and verify that the provider documents backup behaviour.

Disk latency matters more than raw CPU for many small databases; measure it before migrating a busy workload.

Scope: Small relational databases.

Limits and exceptions

No disk-latency measurements are published in this dataset.

Related: “RAM” · “NVMe” · Small database

Updated: 2026-10-25

Question 44

Can a VPS run multiple containers reliably?

Accepted Answer

Yes, if memory is sized for the sum of container limits plus the host overhead, and the virtualization type supports the networking and storage features you need. KVM is the safer default for Docker and nested networking. Track memory pressure over time; containers that idle small can spike during builds or scheduled jobs.

Question 45

What should I check before renting a game server VPS?

Accepted Answer

Check single-thread CPU performance expectations, the plan's port speed and traffic policy, the region's latency to your players, and whether the provider's terms permit game servers. Memory is often the first limit hit. Test with representative player counts before committing to an annual term.

Question 46

How should I compare two VPS providers fairly?

Accepted Answer

Compare the same dimensions on both sides: the renewal-aware 12-month cost, the exact resources, storage type, transfer and port speed, address inclusion, backup terms, locations and evidence dates. Normalize units and currencies before ranking. Then read the provider's limitations, not only the marketing table. A difference in a field you do not need is not a reason to pay more.

Question 47

Why doesn't VPSTab publish a single "best VPS provider" ranking?

Accepted Answer

A single ranking would have to combine price, resources, policy, support and performance into one number. This dataset has no independent performance, support or reliability measurements, so such a number would be invented. Instead, this site publishes evidence-scored assessments dimension by dimension, withholds overall scores where evidence is missing, and lets the recommendation engine apply explicit scenario weights.

Question 48

Are small providers riskier than large ones?

Accepted Answer

Small providers can offer better prices and direct support, but they may have fewer regions, less published documentation and higher business-continuity risk. Large providers are more documented but often bundle fewer resources per euro.

Judge the evidence you can verify: terms, backup policy, published status and how clearly the provider states limitations. Size is not a substitute for either.

Scope: Provider risk assessment.

Limits and exceptions

This dataset does not score financial stability or support quality.

Related: “Evidence class”

Updated: 2026-10-25

Question 49

What do the "data gap" counts on a provider page mean?

Accepted Answer

They show how many of that provider's recorded offers leave a decision-relevant field unpublished, such as IPv6 inclusion, backup availability or tax treatment. Unknown is not "no" and not zero. A high count does not mean the provider is bad; it means you must verify more at checkout before relying on those fields.

Question 50

What do the evidence classes mean?

Accepted Answer

Each fact carries a provenance label: provider official (a page or document published by the provider), provider API, independent test, third party, community or editorial. The label tells you who is making the claim. Provider-stated hardware or network figures are not independent measurements; this site marks them as provider claims rather than converting them into performance rankings.

Question 51

Why do some reviews show no overall score?

Accepted Answer

An overall score requires evidence across all decision-critical dimensions. When performance, support and reliability measurements are missing, a total score would be an average of guesses, so it is withheld. Dimension scores that are shown describe evidence coverage for that dimension, not a universal product rating.

Question 52

Can I trust a provider's claim that its network is faster?

Accepted Answer

Treat speed claims as marketing until you can reproduce them. Providers often describe hardware (CPU model, port speed, network fabric) and imply that the hardware is faster than a competitor's, but the same claim can be true on paper and irrelevant to your workload. Independent measurement on the plan you intend to buy is the only evidence that covers your region, routing and load pattern. This dataset publishes no such measurements, so it never converts a hardware statement into a performance ranking.

Question 53

How does VPSTab handle stale prices?

Accepted Answer

Every offer carries an observation date and a verification date. The recommendation engine excludes offers whose evidence is older than the freshness window and explains the exclusion instead of ranking them silently. For historical context, the plan page keeps an offer-history record so an older price is visible as history, never as the current price.

Question 54

Why does "unknown" sometimes mean I cannot get a recommendation?

Accepted Answer

A hard requirement is a condition you must be able to verify. If you require IPv6 and the provider does not state whether IPv6 is included, the plan has not met the requirement; it has not failed it either. The engine excludes unknown-critical fields from matches rather than assuming yes, and the interface shows which field blocked the plan.

Question 55

What should I verify on the provider's checkout page?

Accepted Answer

Confirm the exact plan name and size, the region and datacenter, the billing term, the currency, the renewal price, whether a primary IPv4 address is included, and any setup fee or mandatory add-on. These are the fields most often different from the marketing page.

Keep a screenshot or record of the checkout total; if it differs from the recorded price, update the observation instead of forcing the old data to fit.

Scope: All purchases.

Limits and exceptions

Checkout availability can change between sessions.

Related: Guides: The pre-purchase VPS checklist · Guides: The true monthly cost of a VPS · “Billing period” · “Tax scope”

Updated: 2026-10-25

Question 56

How do I confirm that a region has stock before paying?

Accepted Answer

Start an order for the exact size and region and look for a stock error before entering payment details. Region catalogs and API availability lists are hints; only the order flow confirms a specific size at a specific facility. If a plan is out of stock, record that observation and check whether a nearby region or size is acceptable.

Question 57

Should backup terms affect my choice before I buy?

Accepted Answer

Yes, if the data cannot be rebuilt. Provider backup schedules, retention and prices differ enough to change the total cost and the recovery plan. Some providers include weekly backups; others sell a percentage add-on. Decide whether the provider's backup alone meets your targets or whether you also need an off-provider copy, then include that cost in the comparison.

Question 58

Is it reasonable to buy a short trial before a long commitment?

Accepted Answer

Yes. A short monthly term lets you test provisioning, region availability, disk performance with your own workload and support response before paying for a year. The monthly premium is often smaller than the risk of a wrong annual commitment.

Use the trial to run the same checks you would run in production, not just to confirm that the instance boots.

Scope: Long-term commitments.

Limits and exceptions

Refund terms may differ between monthly and annual purchases.

Related: “Billing period” · “Renewal price”

Updated: 2026-10-25

Question 59

How do I move a service to a different VPS provider?

Accepted Answer

Prepare the destination first: provision the new instance, install the same stack, restore data from a tested backup, and verify the service privately before switching DNS. Keep the source running during the overlap.

After the switch, monitor errors and latency for a few days before cancelling the old instance; keep an off-provider backup from before the move.

Scope: Single-instance migrations.

Limits and exceptions

Large datasets require planning for transfer time and cost.

Related: Guides: Backup and restore planning for a single VPS

Updated: 2026-10-25

Question 60

What happens to DNS when the IP address changes?

Accepted Answer

Update A and AAAA records to the new addresses, lower the TTL before the move so stale records expire quickly, and wait for propagation before cancelling the old service. If you run mail, update SPF, DKIM, DMARC and reverse DNS as well.

Keep the old instance reachable until the new one has served traffic without errors.

Scope: DNS-based migrations.

Limits and exceptions

Propagation time depends on the previous TTL and resolver caches.

Related: “IPv4” · “IPv6”

Updated: 2026-10-25

Question 61

Can I migrate without downtime?

Accepted Answer

For stateless services, yes: run both instances behind DNS or a load balancer and drain traffic. For stateful services, use replication and a controlled cutover, or accept a short maintenance window.

A single-instance migration with a database usually requires a write pause; plan it explicitly instead of discovering it during the move.

Scope: Migration planning.

Limits and exceptions

Provider-to-provider private networking is not available in most cases.

Related: Self-hosted applications · Small database

Updated: 2026-10-25

Question 62

How do I move large data volumes off a VPS?

Accepted Answer

Compress first, then transfer over a fast link in resumable chunks. Watch the source provider's egress allowance and the destination's inbound policy; a large transfer can consume a monthly traffic cap. For very large datasets, a provider-supported export or object-storage staging area can be cheaper than transferring directly between instances.

Question 63

Who is responsible for patching a VPS?

Accepted Answer

You are. The provider supplies the virtualization layer and often a base image; the operating system packages, application dependencies and configuration are your responsibility unless you buy a managed service.

Schedule unattended upgrades or a maintenance routine, and reboot when the kernel changes.

Scope: Self-managed plans.

Limits and exceptions

Managed offerings shift some responsibility back to the provider.

Related: Self-hosted applications

Updated: 2026-10-25

Question 64

What are the basics of securing SSH access?

Accepted Answer

Use key-based authentication, disable password login and root login, restrict the source addresses where possible and keep the SSH daemon updated. A non-standard port reduces noise but is not a security control by itself.

Keep an out-of-band recovery path, such as provider console access or a second key, before changing SSH configuration.

Scope: Linux servers.

Limits and exceptions

Provider console access can be limited or unavailable on some plans.

Related: “Unknown (data gap)”

Updated: 2026-10-25

VPS questions and answers

VPS basics

Budget and long-term cost

Promotions, renewals and billing

CPU, architecture and virtualization

Memory, storage and limits

Traffic, bandwidth and networking

Locations and latency

IPv4, IPv6 and addresses

Backups, snapshots and recovery

Refunds, cancellation and policies

Workloads and plan sizing

Comparing providers

Data sources and review credibility

Before you buy

Migration and operations

Security and reliability