As a scheduling component, PD maintains meta information about the entire TiKV cluster. This includes not only the region topology but also things like store read/write traffic, CPU usage, etc. This information is not only important for PD to make scheduling decisions, but also for TiKV to help make better decisions for certain functions. For example, Load Base Split is a key functionality for TiKV to split the hot region proactively in order to make PD be able to spread them out. With this, the whole TiKV cluster could have a more balanced load. However, Load Base Split works independently on each TiKV which lacks a more holistic perspective. This may lead to problems such as not being split evenly enough, over-split, etc. To solve this, providing feedback functionality for TiKV is crucial. With PD's guidance and feedback, TiKV could make "smarter" decisions.