Data distribution patterns When some of the columns (attributes) of a relation are placed at different sites while rows may remain aligned by primary key, this is called:

Introduction / Context:
Distributed databases often split tables to improve locality, security, and performance. Two primary strategies are horizontal partitioning (by rows) and vertical partitioning (by columns). Knowing which method is used helps predict query routing, join costs, and storage design.

Given Data / Assumptions:

• Some columns of a relation are stored at different sites.
• A common key (typically the primary key) exists for reconstructing the full row via joins.
• Goal: identify the correct partitioning terminology.

Concept / Approach:

Vertical partitioning divides a table by columns. Each fragment contains the key plus a subset of attributes to enable reconstruction (lossless decomposition). In contrast, horizontal partitioning divides by rows (tuples), typically using predicates or hashing. Replication duplicates data across sites rather than splitting it. A hybrid approach uses both row and column fragmentation simultaneously.

Step-by-Step Solution:

Verification / Alternative check:

Textbooks define vertical partitioning as attribute-wise distribution with the key repeated; horizontal partitioning distributes tuples based on predicates or hashing across sites.

Why Other Options Are Wrong:

• Replication: copies the same data at multiple sites.
• Horizontal partitioning: distributes rows, not columns.
• Hybrid: both are used; the stem mentions only columns.
• Shuffling: vague, not a standard term for schema design.

Common Pitfalls:

• Omitting the key in vertical fragments, making lossless joins impossible.
• Confusing columnar storage formats with vertical partitioning (they are related but distinct concepts).

Final Answer:

Vertical partitioning.