Currently in the design doc, we implement a hash table with a fixed size derived from the WT cache size. Each bucket in the table is protected by a lock that guards insert, update, and remove operations. Within each bucket, entries are stored in a doubly linked list, with each entry representing a page cache.

The concern is that when multiple entries collide into the same bucket, operations on that bucket will contend on the same lock. Since page read is on a critical path, this level of contention may not be acceptable, especially if the hash table size is not well tuned and collisions become frequent.

This ticket is to implement a POC and run perf tests to verify whether the current page cache hash table design is causing performance degradations, and propose workaround if it does.