THEORY EXAMINATION (SEM–VI) 2016-17 DISTRIBUTED DATABASE

DISTRIBUTED DATABASE – NCS067

B.Tech (SEM VI) | Section-wise Solved Answers

SECTION – A

(10 × 2 = 20 marks)

(a) Cascadeless schedules

A cascadeless schedule is one in which transactions read only committed data. This avoids cascading rollbacks because no transaction depends on uncommitted updates of another transaction.

(b) Serializability and its testing

Serializability ensures that the outcome of concurrent transactions is equivalent to some serial execution. It is tested using a precedence (serialization) graph. If the graph has no cycle, the schedule is serializable.

(c) Strict 2PL vs Rigorous 2PL

In strict two-phase locking, exclusive locks are released only after commit, while shared locks may be released earlier.
In rigorous two-phase locking, both shared and exclusive locks are held until commit, ensuring strict serializability.

(d) Timestamps associated with each data item

Each data item maintains a read timestamp (RTS) and write timestamp (WTS) to track the most recent read and write operations.

(e) Replication

Replication is the process of maintaining multiple copies of data at different sites to improve availability, reliability, and performance.

(f) Fragmentation and its types

Fragmentation divides a database into smaller pieces.
Types include horizontal fragmentation, vertical fragmentation, and hybrid fragmentation.

(g) Orphan concept in distributed database

An orphan transaction occurs when a transaction continues execution even though its parent transaction has failed, leading to inconsistency.

(h) Database recovery

Recovery restores the database to a consistent state after failure using log-based recovery, checkpointing, and rollback/redo operations.

(i) Distributed vs Replicated database

A distributed database stores data across multiple sites.
A replicated database stores multiple copies of the same data at different sites.

(j) Distributed deadlock detection

Distributed deadlock detection identifies cycles in the global wait-for graph across multiple sites to detect deadlocks involving distributed transactions.

SECTION – B

(Attempt any five – 5 × 10 = 50 marks)

(a) Conflict vs View Serializability

Conflict serializability is based on swapping non-conflicting operations and is easier to test using precedence graphs.
View serializability considers read-from relationships and final writes and is more general but difficult to test.

(b) Timestamp-based concurrency control

Transactions are ordered using timestamps. Operations violating timestamp order are aborted. This method avoids deadlocks but may cause frequent rollbacks.

(c) Purpose of Two-Phase Commit (2PC) Protocol

2PC ensures atomicity of distributed transactions.
It guarantees that either all participating sites commit the transaction or all abort.

(d) Two phases of 2PC

Phase 1 – Voting phase: Coordinator asks participants to prepare.
Phase 2 – Decision phase: Coordinator decides commit or abort based on votes and informs participants.

(e) Cost-based query optimization

This approach selects the execution plan with minimum communication and processing cost by estimating data transfer, CPU usage, and response time.

(f) Recovery in message passing systems

Failures can lead to inconsistent states where messages are lost or duplicated. Logging and checkpointing help restore consistency.

(g) Multiversion schemes

Multiversion concurrency control maintains multiple versions of data items. It improves read performance and reduces conflicts between read and write operations.

(h) Query optimization algorithms

Algorithms include dynamic programming, heuristic-based optimization, and cost-based optimization to minimize distributed query execution cost.

SECTION – C

(Attempt any two – 2 × 15 = 30 marks)

(3) Transaction management in distributed databases

Distributed transaction management ensures atomicity, consistency, isolation, and durability across sites.
Homogeneous DB: Same DBMS and data model at all sites.
Heterogeneous DB: Different DBMSs and data models at different sites.

(4) Traditional recovery techniques

Traditional recovery uses write-ahead logging, undo/redo operations, and checkpoints.
Example: If a system crashes after logging but before commit, redo is applied; if before commit, undo is applied.

(5) Objectives of distributed query processing

The objectives are to minimize data transfer, reduce response time, optimize resource usage, and ensure correctness of query results in a distributed environment.