Hosting CPU, RAM and I/O Limits Explained

A website can run out of processing resources while plenty of disk space remains. Storage tells you how much data an account can hold; CPU, memory and I/O describe different parts of the work needed to serve that data.

Read each limit separately

CPU concerns computation, memory holds data used by running tasks, and disk I/O concerns reads and writes. Process and concurrency limits can also affect how many tasks run together. These measurements are related, but increasing one allowance does not automatically remove every bottleneck.

The panel’s graphs are more useful when matched to an exact event. Note whether a limit coincides with a backup, catalogue import, bot activity or genuine customer traffic. An isolated peak is not the same as sustained saturation during normal business hours.

Collect a short incident record

Record the affected URL, time and time zone, response or error message, and the task being performed. Include the relevant resource graph, with private customer details removed. This gives support a reproducible problem rather than a general report that “hosting is slow.”

For example, image regeneration may briefly consume CPU while the website otherwise behaves normally. A search function that repeatedly exhausts resources during ordinary use needs a different review: query design, indexing and concurrency should be investigated.

Match the change to the evidence

Schedule heavy maintenance away from peak traffic where practical. Review expensive queries and unnecessary background work. If the normal workload still reaches a measured limit, compare packages that increase that particular resource and clarify what happens during an upgrade.

• Compare charts from the same time interval.
• Distinguish a per-task memory setting from an account-wide allowance.
• Ask how the provider defines concurrency on its stack.
• Verify the result after a change, rather than assuming an upgrade fixed it.

Check the published resources in OrienHost hosting plans, and use the incident record when asking which package fits the measured workload.