The performance of WiredTiger applications, especially heavily-threaded applications can be dominated by memory allocation because the WiredTiger engine frees and re-allocates memory as part of many queries. Replacing the system's malloc implementation with one that has better threaded performance (for example, Google's tcmalloc, or FreeBSD's jemalloc), can dramatically improve throughput.

As different memory allocators have different overhead and different workloads will have different heap allocation sizes and patterns, applications may need to set their allocator overhead using the cache_overhead configuration to the wiredtiger_open:: call.