CopyonWrite

Copy-on-write (COW) is a resource management strategy in which multiple callers can share a single instance of data and only incur the cost of copying when a modification is necessary. Under COW, reads can access the shared data without duplication, while writes trigger a duplication so that the writer receives an isolated copy. This approach aims to reduce unnecessary copying and improve performance by deferring work until it is actually needed.

In memory management, COW is famously used during process creation in many operating systems. When a process forks, the parent and child initially share the same physical memory pages. These pages are marked as read-only, and a reference count tracks how many processes reference each page. If either process writes to a shared page, a page fault occurs; the kernel then creates a private copy of the page for the writing process and updates the page table. Subsequent writes affect only the private copy, while reads continue to reference the shared data until a copy is required.

In file systems and storage, COW enables efficient snapshots and atomic updates. Instead of overwriting blocks in place, new blocks are allocated for changes and metadata updates are performed atomically, preserving previous versions. File systems such as Btrfs, ZFS, and APFS use COW to implement snapshots, rollbacks, and crash-consistent updates with reduced risk of data corruption and without duplicating unchanged data.

Applications beyond file systems include persistent data structures in programming, where structural sharing allows efficient versioning without copying entire structures. While COW offers performance and memory advantages, it introduces overhead for reference tracking and potentially increased write amplification in certain workloads.