> ## Documentation Index
> Fetch the complete documentation index at: https://docs.condense.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Glossary

> The words used in the Autoscaling pages: requests and limits, the recommended size, replicas, restart, service and the Insights tab.

* **Requests for CPU, Requests for RAM.** What your connector reserves. Work is placed using these figures, and your invoice is calculated from them. Reserved capacity is held for your connector whether or not it uses it.
* **Limits for CPU, Limits for RAM.** The cap on what your connector may use. Past its memory cap, the connector is stopped. Past its CPU cap, it is held back and runs slower.
* **The recommended size.** The memory or CPU figure Condense recommends for your connector, worked out from what it has really used over the last eight days. It is the figure a recommendation offers.
* **Replica.** One running copy of your connector. Replica scaling changes how many copies there are. Resource sizing changes how big each copy is.
* **Restart.** Your connector stopping and starting again on new figures. It is the cost of accepting a recommendation, and you are told about it before you commit.
* **Service.** The word the screens use where these pages say connector, for example "Nothing to Change on this Service" and "restart this service". It means the connector you have open.
* **The Insights tab.** Where recommendations for a connector appear, beside its Observability, Alerts, Logs and Environment Variables tabs.
* **VPA.** The Kubernetes Vertical Pod Autoscaler, which Condense runs in a mode that only advises. It measures what your connector uses. The screens refer to it by name in places.

> [card] [Back to Autoscaling](https://docs.condense.io/condense/v3.0.0/autoscaling/overview.md)
>
> The overview of what Autoscaling does for you.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.