Shared or dedicated CPU: Choose from the load profile

Standard as starting point

The WordPane catalog features standard plans with shared vCPU. They can be evaluated for websites, stores, and applications whose consumption is compatible with the available features. The behavior depends on the application and the provider; the name of the plan does not define the result alone.

Use the start of the operation to measure CPU, memory, disk, and peak times. Also register scheduled tasks such as import, report, and image processing. These tasks may explain a load that does not appear in total visits.

When evaluating dedicated CPU

Dedicated CPU deserves evaluation when processing has an important participation and you search for a more predictable profile of resources. Before the change, check if the bottleneck is really in processing, not in memory, database or network.

A heavy report may require optimization or background execution beyond a different instance. There is no universal performance guarantee: test your flow with a representative load.

AI does not always mean CPU anymore

If the server calls an external template, a relevant part of the time can be waited by network. If it prepares documents, transforms files or executes local algorithms, the demand changes. Identify where the work happens before choosing the plan.

Running an AI model on the server itself requires evaluating specific requirements and does not automatically resolve with dedicated CPU. Inform model size, need for GPU and volume of use when requesting a proposal.

Distinguish brief peaks from sustained processing

An institutional website can concentrate work on brief moments and remain idle between visits. A file conversion or a batch of reports can keep processing for long periods. These profiles have different predictability and capacity needs.

Dedicated CPU may be appropriate when sustained load and processing variability matter, but exact conditions depend on the provider and plan. The term does not guarantee latency, isolation of all resources or freedom from network and disk contention.

Compare plans with an equivalent scenario

Use the same dataset, application version and access pattern. Record CPU usage, memory, operation duration and errors. If one step uses only one core, doubling the vCPU count may not halve its duration.

If the wait is in a remote database or an AI provider, adding local CPU may have little effect. If memory is exhausted and swap usage is high, a different processor class alone will not solve the problem. Interpret the whole request path.

Example: automations during the service hours

A company may have a public website and tasks processing many records. When they compete for resources, the site may slow down while the tasks run. Evaluate scheduling, limits and process separation before expanding everything.

Test batch jobs and visits at the same time. Check that job duration and response times stay within business targets. The result guides capacity choices and helps decide whether these services should share a server.

Decide with total cost and operating margin

Consider region, storage, backup, additional features and contracted management. A monthly reference may derive from hourly billing and vary with period and components. Please refer to the current WordPane catalog and plan availability at the time of hiring.

Also define when re-evaluate: consistent queue increase, latency above the target or growth of the data set. This criterion avoids expanding by anxiety and avoids waiting for the service to fail to recognize that the profile has changed.

Compare cost with real use

• Separate mean consumption and recurrent peaks.
• Observe memory pressure and available space.
• Include external API costs in the budget.
• Check current resources and values per region in the catalog.
• Plan tests and possible interruptions before resize.