JAWAHARLAL NEHRU TECHNOLOGICAL UNIVERSITY HYDERABAD B.Tech III DISTRIBUTED DATABASES Previous Question Paper (September - 2021 - R18)

The design of a distributed database system must balance local autonomy with global coordination. Key requirements include: Transparency (location, fragmentation, and replication transparency so the system appears as a centralized database), Reliability (the system must continue to function despite individual node failures), Scalability (the ability to handle increased data volumes and user loads by adding nodes), and Performance (minimizing communication overhead and latency).

The primary challenges include: Distributed Query Processing (optimizing queries across multiple nodes to minimize data transfer), Distributed Concurrency Control (maintaining consistency across nodes using protocols like two-phase locking or timestamp ordering), Data Consistency and Replication (ensuring all copies of data reflect the same state), and Heterogeneity (managing differences in hardware, operating systems, and DBMS software across nodes).

Fragmented relation simplification is the process of transforming a global query into an optimized query on fragments. This involves several steps:

1. Selection Simplification: Applying selection predicates to horizontal fragments. If a predicate on a fragment is always false, the fragment is pruned from the query.
2. Projection Simplification: Removing unnecessary attributes from the query. If a projection is performed on a vertical fragment, only those fragments containing the required attributes are accessed.
3. Join Simplification: Utilizing the properties of fragmentation to eliminate joins. For example, if two relations are partitioned based on the same join attribute (co-partitioning), the join can be performed locally at each site.
4. Redundancy Elimination: Identifying and removing redundant operations or redundant fragments that do not contribute to the final result set, often using algebraic equivalences to rewrite the query tree into a more efficient form.