A useful VPS budget starts with the work the server must do. A cached website, an API waiting on another service and an automation that processes large files can have very different resource needs. The plan name or a visitor estimate cannot describe those differences. Write down the application, its data and the processes that will run together before comparing prices.
ServerQ offers KVM VPS plans in Istanbul and New York. Each plan allocates vCPU, RAM and disk to the customer; this does not promise a separate physical CPU core. The two locations have their own configurations. Use this guide to establish a starting requirement, compare actual offers and decide which measurements will tell you when to change it.
What this guide helps you do
- Separate memory, CPU, storage and network requirements.
- Choose a pilot plan with explicit assumptions and operating headroom.
- Compare the location, billing period and recovery needs before ordering.
Developers, small businesses and teams choosing a self-managed VPS for a website, API or automation.
Start with the processes, not a plan label
List the web server, application workers, database, scheduled jobs, logs and monitoring tools. They may all share the VPS. Include temporary tasks such as image conversion, updates and backup compression: a server that fits the quiet period can still run out of memory during maintenance.
For a new project, choose a representative pilot rather than claiming a production capacity in advance. Use realistic data sizes and the busiest workflow you expect. Record the assumption behind the first plan so that an upgrade later responds to a measured constraint.
- 01Describe the workload
- 02Choose a starting allocation
- 03Measure real behaviour
- 04Change the limiting resource
A larger plan should solve a named problem, such as memory pressure or a growing job queue.
Read each resource as a different constraint
| Resource | What it provides | What to observe |
|---|---|---|
| RAM | Working space for the OS, application, database and caches. | Available memory, out-of-memory events, swap activity and peak worker use. |
| vCPU | Virtual processing capacity for concurrent or CPU-heavy work. | Sustained CPU demand, job completion time and request latency. |
| Disk | Space for application data, logs, media and temporary work. | Free bytes, free inodes, growth rate and I/O waiting. |
| Network | Data transfer and the route to users and dependencies. | Monthly transfer, connection time, external-service delays and traffic terms. |
More vCPU does not make an oversized database fit into memory. More disk space does not guarantee faster queries. NVMe describes the storage technology, while the performance your application sees also depends on its access pattern and the rest of the system. Compare the actual plan resources and test the application outcome.
The published port value describes the hypervisor host connection. It is not a dedicated per-customer throughput promise. Istanbul offers use the transfer allowance shown on the plan. USA offers have no fixed traffic quota and remain subject to fair use; that is not unlimited line-rate capacity.
Use workload presets as a starting question
| Workload | Likely first concern | What changes the requirement |
|---|---|---|
| Static site or reverse proxy | Connection handling, logs and a small OS/application baseline. | TLS work, uncached requests, file size and monthly transfer. |
| WordPress with a local database | PHP worker memory, database working set and disk. | Plugins, dynamic checkout/account pages and scheduled jobs. |
| Node.js API | Application memory, concurrent tasks and dependencies. | Payload size, request duration, background jobs and database queries. |
| n8n automation | Peak memory per execution and overlapping workflows. | Large files, Code nodes, execution history and API rate limits. |
The plan finder lets you set minimum vCPU, RAM and storage and filter by a monthly budget. Its presets help begin the comparison; they are not measurements of your application. Leave room for the operating system, the database if it is local, and tasks that overlap. If no plan meets the filters, review the assumptions or discuss the requirement instead of treating the nearest result as a fit.
Connect system measurements to an application result
free -h
df -h
df -i
uptime
vmstat 1 5
cat /proc/pressure/memory
cat /proc/pressure/ioCommand availability and kernel interfaces depend on the installed distribution. These checks are snapshots, not long-term monitoring. Linux uses spare memory for cache, so a low “free” value alone is not evidence that the plan is too small. Record the busy period as well as a quiet baseline.
Pair these observations with application error rate, queue delay and a representative response-time percentile. A slow upstream API can leave your CPU idle while requests time out. An inefficient query or missing cache can consume resources that a code or database change would avoid. Change one factor and compare the same workload again.
Compare one region, currency and billing period at a time
Choose Istanbul or New York using the users and services that make up your request path. The RAM, disk and transfer policy of one region must not be copied onto another region’s offer. Read the location label in the product card and again in the portal before completing the order.
Compare an annual total with twelve monthly payments for the same plan, not with one monthly payment. ServerQ applies no additional tax to the listed service price. Paid operating-system or control-panel licences, when selected, are separate additions. Confirm the selected currency and final total in the portal; a display-currency selection on the website does not itself set the portal currency.
Finish with a short, reviewable purchase decision
- The product and portal show the intended location and plan.
- Allocated vCPU, RAM and disk meet the pilot assumptions with operating headroom.
- The selected OS image, any paid licence and the support scope are confirmed.
- The host-port description and the region’s transfer policy are understood.
- A person owns updates, monitoring, backups and recovery.
- The reason for the next review is defined: a resource limit, an error trend or a change in workload.
Keep this decision with the deployment notes. When the workload grows, you can explain which resource needs to change and why. Confirm how a resize or migration will affect the application before scheduling it; a higher specification does not automatically imply a live, interruption-free transition.
Review scope
Resource suggestions are starting assumptions, not guaranteed visitor, order or workflow capacity. ServerQ confirmed the KVM resource allocation and location-specific network scope on 30 September 2026. Compare the current plan and portal order summary before purchasing; application capacity must be measured with your workload.
Official references
Check the project’s current documentation for version and configuration details.
- Linux kernel — Pressure Stall Information (new tab)
How pressure metrics describe resource stalls.
- Ubuntu Server documentation (new tab)
Distribution-specific operating and monitoring guidance; not a promise that every image is offered.