Chunk cache can be configured to allocate cached chunks in RAM or on disk (typically SSD). The current implementation (as of March 2023) uses libmemkind to allocate data on disk – it provides a malloc API, with jemalloc carving the space in an mmapped chunk of disk space. There are bugs in libmemkind and we are unsure of its future. This library was created by Intel to support Optane NVRAM, but this product has been cancelled. 

The purpose of this ticket is to eliminate dependence on libmemkind, for example to adopt WiredTiger block manager to manage space on the disk for the chunk cache.