This topic describes the architecture and storage mechanism of backup files.
Architecture
OceanBase Database backs up data from the follower replicas. The following figure shows the architecture of data backup.
Storage mechanism
In the data backup scenario, the native NFS protocol has poor performance when handling a large number of small files (for example, 2 MB files in the millions). To address this issue, OceanBase Database V4.x introduces a data aggregation storage mechanism. This mechanism packages each backup item (backup item) for storage, thereby reducing the number of small files. A backup item consists of three types of data: SSTable metadata, tablet metadata, and macroblock objects. By default, 2048 backup items (each up to 2 MB) are aggregated in each batch. The number of backup items aggregated in each batch (backup_data_file_size / 2 MB) can be adjusted by the tenant-level parameter backup_data_file_size. The sizes of the aggregated files vary due to the differences in the sizes of the metadata and macroblocks.