Fragmentation of data in distributed database----"Distributed database system and its application"

Data fragmentation, also known as data segmentation, is one of the characteristics of distributed database. In a distributed database, the global database is composed of the logical combination of the local database, whereas the local database is partitioned by some logic of the global database.

In a distributed database, the units in which data is stored are the logical fragments of the data. For relational databases, a logical fragment of a database is part of a relationship. There are three basic methods of data fragmentation, which are realized by the basic operation of relational algebra.

1. Horizontal sharding

   All the meta-ancestors of a global relationship are divided into several disjoint subsets according to certain conditions, and each subset is a logical fragment of the global relationship, or fragment for short. They are obtained by applying selection operations to global relationships and can be restored by performing a merge operation on those fragments.

2. Vertical shards

Divides the attribute set of a global relationship into subsets. In order to get these subsets, the global relationship is projected, and each attribute of the global relationship is required to be mapped to at least one vertical fragment, and each vertical fragment contains the construction of the global relationship. This allows you to restore the global relationship by performing a join operation on those fragments.

3) Mixed shards

is a mixture of the above two methods. The horizontal shards are then vertically fragmented, or the vertical shards are then horizontally fragmented, but their results are not the same.

Distributed Database Shard, is not felt in the current Internet architecture is very familiar with, that is, to solve the database bottleneck vertical split and horizontal split.

Fragmentation of data in distributed database----"Distributed database system and its application"