WiredTiger uses multiversion concurrency control (MVCC).

MVCC is based upon the concept that the database keeps multiple versions of an object so that readers will be able to view consistent data that doesn't change during a read.

In a database, if we have multiple readers accessing data at the same time that writers are modifying the data, we can end up with a case that readers view an inconsistent view of this data. The simplest and easiest way to solve this problem is to block all readers until the writers are done modifying data.

This will of course cause severe performance degradation. MVCC solves this problem by providing a snapshot of the database for each reader. At the time that the read starts, each reader is guaranteed to view data as it was at this point in time.

Any changes made by writers will only be seen by readers after the write has completed, or in database terms, after the transaction is committed.

To achieve this goal, when a write is coming in, updated data will be kept in a separate location on disk and MongoDB will mark the affected document as obsolete. MVCC is said to provide point in time consistent views.

This is equivalent to a read committed isolation level in traditional RDBMS systems.

For every operation, WiredTiger will snapshot our data at the exact moment that it happens and provide a consistent view of application data to the application.

When we write data, WiredTiger will create a snapshot every 2 GB of journal data or 60 seconds, whichever comes first. WiredTiger relies on its built-in journal to recover any data after the latest checkpoint in case of failure.