Best starting shortlist for a lean Linux app: Hostinger KVM 1. Pick KVM 2 when your app, database and overlapping jobs need more headroom; compare UltaHost when management support is the priority. Compare Hostinger VPS plans
Affiliate disclosure: Some links are affiliate links. We may earn a commission on qualifying purchases at no extra cost to you. Read our disclosure .
Based on official pricing and documentation checked November 5, 2026; sizing recommendations are editorial starting points to validate with your workload.
Compare RAM allocations and the full purchase cost
Start by listing everything that will run together. Include the operating system, application, database, queue workers, monitoring and maintenance jobs. Buying for the application alone leaves the rest of the stack competing for the same memory.
| Plan | RAM / vCPU / NVMe | Intro USD monthly equivalent | Initial payment | Renewal monthly equivalent | Suggested fit |
|---|---|---|---|---|---|
| Hostinger KVM 1 | 4 GB / 1 / 50 GB | $6.49, 24 months | $155.76 | $11.99, same term | Lean app |
| Hostinger KVM 2 | 8 GB / 2 / 100 GB | $8.79, 24 months | $210.96 | $14.99, same term | App plus database |
| Hostinger KVM 4 | 16 GB / 4 / 200 GB | $12.99, 24 months | $311.76 | $28.99, same term | Larger concurrent stack |
| UltaHost VPS Basic | 1 GB / 1 / 30 GB | $5.52, displayed two-year offer | Confirm checkout | $5.52, two years | Small managed project |
Sources: Hostinger VPS pricing and resource specifications , checked November 5, 2026; UltaHost VPS Basic pricing and specifications , checked November 5, 2026. Suggested fits are our judgments. Both pages advertise a 30-day money-back guarantee, rather than a free trial. Hostinger requires upfront payment; verify taxes and the final invoice before purchasing.
The cheapest advertised monthly equivalent is not necessarily the smallest commitment. Budget for renewal and any software licenses, external backups or administration you need. Keep the project affordable after the introductory term, especially when a client contract may end before the hosting term does.
Size by workload, not by website count
The following shortlist uses the plan capacities above as buying tiers. These are not vendor minimum requirements or guaranteed traffic limits. Start at the lower suitable tier only when you can observe the application and change the configuration if needed.
| Workload | Starting shortlist | What could justify moving up? |
|---|---|---|
| Cached portfolio or small API | KVM 1 | More processes, heavier requests or a growing local database |
| WordPress with database on the same server | Evaluate KVM 1; shortlist KVM 2 for busier stacks | Dynamic requests, imports and background jobs overlapping |
| Docker app with database and workers | KVM 2 shortlist | Several active services competing for memory |
| n8n with modest text workflows | Evaluate KVM 1 | Large files, bulky execution data or simultaneous workflows |
| Multiple client apps or substantial batch processing | Evaluate KVM 2 or KVM 4 | Observed aggregate peak memory exceeding the smaller plan |
For WordPress, a brochure site with effective caching differs from a store running checkout, search and scheduled imports. Do not buy from a universal visitors-per-day estimate. Reproduce the actions customers actually take, including uncached pages, before choosing the smaller plan.
For containers, count processes and services rather than images. A stopped image consumes storage; a running database or worker needs resources. Our Docker VPS buying guide covers the related storage and operating-cost decisions.
For automation, payload size matters as much as workflow count. n8n documents memory pressure from large data, binary files, certain processing nodes and concurrent executions. It also explains that manual execution makes an additional frontend copy. Source: n8n memory troubleshooting documentation , checked November 5, 2026. Use the self-hosted n8n VPS guide when choosing the wider stack.
A bot calling an external AI API is a different purchase from a server loading model weights locally. Before buying for local inference, inspect the chosen model’s own memory and hardware documentation. Do not infer that an ordinary VPS tier is suitable simply because the product page mentions AI.
Build a memory budget before choosing a tier
Use this planning relationship: required capacity equals baseline services plus simultaneous application demand plus maintenance headroom. It is a method for organizing your observations, not a benchmark formula with fixed constants.
First record the idle baseline after installing your intended stack. Then run a representative task and record the increase. Repeat with the jobs that could realistically overlap. Keep units consistent and distinguish the memory of one process from the memory available across the host.
Do not add every historical maximum if those tasks cannot run together. Conversely, an average can hide a short peak that restarts a worker. Keep both the peak and its context: the input size, concurrency, software version and background work running at the time.
Choose the next tier when the smaller one leaves insufficient room during that realistic overlap. If a batch import can run overnight without competing with customer requests, rescheduling may be cheaper than a permanent upgrade. If it must run during business hours, additional capacity may protect valuable working time.
Who should choose Hostinger - and who should not
Choose Hostinger when the listed resource tiers match your stack and you can own deployment, configuration and application recovery. KVM 1 is the budget candidate for focused work. KVM 2 is my shortlist for a stack that combines several active services. The KVM 1 vs KVM 2 comparison explains the purchase trade-offs in more detail.
Do not choose the larger tier solely because its RAM figure doubles. More memory helps when memory is the constraint; it does not repair a slow external API, inefficient query or CPU-bound job. Nor does paying for a larger server replace a backup and restore process.
Compare UltaHost if its advertised managed service could reduce the administration you need to perform. The entry plan in the table has a much smaller memory allocation, so it is not an equivalent substitute for a larger Hostinger configuration. Ask which tasks and applications the management service covers before paying.
See Hostinger plans for your memory budget
Recognize memory pressure before upgrading
Docker explains that containers have no resource limits by default and that host memory exhaustion can cause processes to be killed. It supports memory limits, while warning that swap is slower than RAM. Source: Docker memory and resource constraint documentation , checked November 5, 2026.
Use host metrics, container metrics and application logs together. A service restart may reflect a container limit even while the server still has available memory. Raising the hosting tier without checking that limit can leave the original failure unresolved.
Likewise, a high used-memory figure alone is not enough to diagnose a shortage. Check available memory, swap activity and the timing of application failures. Investigate repeated out-of-memory events and restarts during normal work. Compare them with CPU utilization and storage activity so you can identify the actual constraint.
If memory keeps growing after each job and does not settle, investigate retention, caching and application behavior. A larger server may postpone a recurring failure rather than solve it. Record what changed after each configuration adjustment so your purchase decision rests on repeatable observations.
How to test it yourself in 15 minutes
Use an already provisioned staging server with your intended application installed and non-sensitive sample data. This is an initial sizing check; production acceptance also needs observation over normal busy periods.
- Record the baseline. Run
free -hfor host memory anddocker stats --no-streamif using containers. Save the time, software versions and services running. - Run normal work. Open uncached application pages, execute a typical workflow or process a representative input. Record errors and completion behavior alongside memory.
- Check realistic overlap. Repeat while a normal background job runs. Keep the workload within systems you control and use the concurrency you actually expect.
- Inspect failures. Review application and system logs for memory exhaustion or restarts. Check whether a container’s configured limit, host capacity or another resource caused the problem.
- Make a decision. Keep the smaller tier when representative work remains stable with spare capacity. Tune expensive tasks first, then move up if repeatable memory pressure remains.
After the short check, include scheduled jobs, deployment and backup windows in a longer observation period. Keep a current off-server backup and rehearse restoration before migrating client work. Save the results so a future upgrade has a clear reason and a before-and-after comparison.
For private file sync, our Nextcloud VPS buying guide combines the RAM decision with usable storage, backup recovery and the full hosting bill.
FAQ
Is 4 GB RAM enough for a VPS?
It is a reasonable tier to evaluate for a lean application, cached site or modest automation stack. The answer depends on your database, workers and busiest overlapping tasks. Validate those together before putting customer work on the server.
Should I buy 8 GB instead of 4 GB?
Shortlist the larger tier when several active services share the host or realistic peaks leave little available memory. If the smaller configuration stays stable and has headroom, the extra recurring cost may be unnecessary.
Does Docker require a larger VPS?
Choose for the applications inside the containers. Set and inspect resource limits so an individual service cannot consume the entire host. Container count alone does not tell you how much RAM the workload needs.
Can swap replace more RAM?
Treat swap as a buffer, not a promise of equivalent performance. If routine work depends on heavy swapping and becomes unresponsive, reduce memory demand or evaluate a larger tier. Check the application’s behavior before deciding.
Why does n8n fail even with a small number of workflows?
A small workflow count can still involve large payloads or overlapping executions. Inspect execution data, binary files and processing steps. Reducing batch sizes may help before you increase the hosting bill.
Will extra RAM make my VPS faster?
It can help when memory pressure is causing the slowdown. It will not remove every CPU, disk, network or application bottleneck. Buy the resource that addresses the observed problem and compare the same workload afterward.
Choose a lean tier for focused work and pay for more headroom when overlapping tasks justify it. Compare Hostinger plans against your memory budget and renewal cost.