Database Service Layer

From the technical perspective, a storage engine requires some infrastructure components.

Concurrency: The overhead of a storage engine can be reduced by properly employing locks, so as to improve overall performance. In addition, it provides functions such as multi-version concurrency control and snapshot reading.

Transaction: All transactions must meet the ACID requirements and their statuses can be queried.

Memory cache: Typically, storage engines cache indexes and data when accessing them. You can directly process common data in the cache pool, which facilitates the handling speed.

Checkpoint: Though storage engines are different, they all support incremental checkpoint/double write and full checkpoint/full page write. For different applications, you can select incremental checkpoint/double write or full checkpoint/full page write based on different conditions, which is transparent to storage engines.

Log: GaussDB Kernel uses physical logs. The write, transmission, and replay operations of physical logs are transparent to the storage engine.