Distributed DBMS Architectures

Distributed DBMS Architectures

• DDBMS architectures depend on three parameters: distribution, autonomy, and heterogeneity.
• Distribution refers to the physical distribution of data across different sites.
• Autonomy indicates the distribution of control in the database system and the independence of each DBMS.
• Heterogeneity refers to the uniformity or dissimilarity of data models, system components, and databases.

Top-Down Classical DDBMS Architecture

• It contains a Global Schema, Fragmentation Schema, and Allocation Schema.

DDBMS Schemas

• Global Schema is a set of global relations as if the database were not distributed.
• Fragmentation Schema splits a global relation into non-overlapping logical fragments, with a 1:n mapping from relation R to fragments Ri.
• Allocation Schema maps fragments to sites in a 1:1 or 1:n (redundant) manner.
• All fragments corresponding to the same relation R at a site j constitute the physical image Rj.
• A copy of a fragment is denoted by Rji.
• Local Mapping Schema maps physical images to physical objects that local DBMSs manipulate.

Design Alternatives for Tables in a DDBMS

• Non-replicated and non-fragmented
• Fully replicated
• Partially replicated
• Fragmented
• Mixed

Non-replicated & Non-fragmented

• Different tables are placed at different sites.
• Suited to database systems with a low percentage of queries needing to join information in different tables.
• Data is placed in close proximity to the site where it's used most.
• Communication costs can be reduced when an appropriate distribution strategy is adopted.

Fully Replicated

• Each site stores a copy of all database tables.
• Queries are fast, with negligible communication costs.
• Massive redundancy in data incurs high costs during updates.
• It is best suited for systems requiring lots of queries with low DB updates.

Partially Replicated

• Copies of tables or portions are stored at different sites.
• Distributing tables according to the frequency of access accounts for the fact that access frequencies vary across sites.
• The number of copies depends on access query frequency and the site generating the queries.

Fragmented

• A table is divided into fragments or partitions, stored at different sites.
• Useful where all data stored in a table is seldom required at a given site.
• Providing better disaster recovery and increasing parallelism.
• There is only one copy of each fragment, no redundant data.
• The fragmentation techniques are vertical, horizontal, and hybrid.

Mixed Distribution

• Combines fragmentation and partial replication.
• Tables are initially fragmented, horizontal or vertical, and the fragments are partially replicated across sites.
• It occurs according to the frequency of accessing the fragments.

Fragmentation

• It is a design technique that divides a class or single relation into partitions.
• A combination of the partitions provides the original database without any information loss.
• It reduces the amount of irrelevant data accessed, reducing disk accesses.
• Include horizontal, vertical, and hybrid/mixed types.

Rules for Data Fragmentation

• Completeness: All data of the global relation must be mapped into the fragments.
• Reconstruction: It should always be possible to reconstruct each global relation from its fragments.
• Disjointedness: Fragments are disjoint, to explicitly control data replication at the allocation level.

Types of Data Fragmentation

• Vertical Fragmentation: Projection on relation (subset of attributes), reconstruction by join, and no tuple migration for updates.
• Horizontal Fragmentation: Selection on relation (subset of tuples), reconstruction by union, and possible tuple migration for updates.
• Mixed Fragmentation: A fragment is a Select-Project query on relation.

Vertical Fragmentation

• Subdivides a database table vertically into groups; fragments are obtained by projecting the global relation over each group.
• Vertical fragmentation splits a single relation R into sub-relations.
• Relations are in grouping with attributes and frequently accessed by queries
• Projection builds vertical fragments.
• The original relation is reconstructed by joining the fragments.

Horizontal Fragmentation

• It divides a relation (table) horizontally into groups to create subsets of tables.

Mixed/Hybrid Fragmentation

• Combination of horizontal and vertical fragmentations.
• A complex type, as both horizontal and vertical fragmentations occur.
• The original relation is obtained by union or join operations.