Distributed Databases

• Data in distributed databases is stored across various physical locations interconnected by a network.
• The system functions as a single, logical database.
• Benefits include improved scalability, availability, and fault tolerance.
• Data partitioning and replication across sites enhance performance through local access.
• The systems are highly resilient because operation continues even if some sites fail.
• Managing these databases is more complex than centralized ones.
• Data consistency, concurrency control, and transaction management contribute to this complexity.

Architecture

• Homogeneous setups feature identical software and schema across all sites.
• Conversely, heterogeneous setups involve different software and schema at different sites.
• Federated database systems integrate multiple, autonomous databases.
• In federated systems, each database retains its autonomy while participating in a global schema.

Data Distribution

• Distribution can be achieved through fragmentation, replication, or both.
• Fragmentation divides data into logical pieces.
• Tables are split row-wise in horizontal fragmentation.
• Tables are split column-wise in vertical fragmentation.
• Replication involves creating multiple data copies at different sites.
• Replication improves availability and read performance.
• Data replication necessitates careful management to maintain consistency across all replicas.

Distributed Query Processing

• This process optimizes and executes queries across multiple sites.
• The primary goal is to minimize data transfer and execution time.
• Query decomposition breaks down queries into subqueries executable at different sites.
• Data localization moves relevant data to the site of query processing.
• Optimization identifies the most efficient execution plan.
• Execution plan considers data location, network bandwidth, and processing costs.

Distributed Transaction Management

• Transactions in these systems involve operations across multiple sites.
• Distributed transaction management ensures atomicity, consistency, isolation, and durability (ACID properties).
• Achieving ACID properties in a distributed environment is challenging due to network latency and potential site failures.

Two-Phase Commit (2PC)

• 2PC is a widely used protocol which ensures atomicity in distributed transactions.
• 2PC involves a coordinator and multiple participants.
• The coordinator manages the transaction’s global state.
• Participants are the sites involved in the transaction.
• Phase 1: Prepare Phase, the coordinator asks all participants to prepare to commit.
• Each participant performs its part of the transaction.
• Participants then signal their readiness (or refusal) to commit to the coordinator.
• Phase 2: Commit Phase which occurs if all participants are ready, the coordinator sends a commit message to all participants.
• If any participant refuses, the coordinator sends an abort message.
• Participants then either commit or rollback their changes accordingly.
• 2PC ensures that all sites either commit or abort the transaction together.
• 2PC can suffer from blocking if the coordinator fails.
• If a participant is unsure of the coordinator’s decision, it may be blocked indefinitely.

Three-Phase Commit (3PC)

• 3PC is designed to address the blocking problem of 2PC.
• 3PC adds an extra phase to provide fault tolerance.
• Phase 1: Prepare Phase, the coordinator sends a prepare message to all participants.
• Phase 2: Pre-Commit Phase, where if participants are ready, they acknowledge the coordinator.
• The coordinator sends a pre-commit message.
• Phase 3: Commit Phase, the coordinator sends a commit message, and participants commit the transaction.
• 3PC reduces the blocking window compared to 2PC, but is more complex to implement.

Concurrency Control

• Concurrency control manages simultaneous access to shared data to prevent interference between transactions.
• Locking protocols are used to serialize access to data.
• Two-phase locking (2PL) is a common concurrency control protocol.
• Timestamping assigns timestamps to transactions.
• Timestamp ordering ensures that transactions are executed in timestamp order.
• Distributed deadlock detection is required to handle deadlocks spanning multiple sites.
• Deadlock detection can be centralized or distributed.

Data Consistency

• Maintaining data consistency across multiple sites is crucial.
• Replication introduces challenges in keeping replicas synchronized.
• Consistency models define the level of consistency guaranteed.
• Strict consistency requires that all replicas are immediately updated.
• Eventual consistency allows temporary inconsistencies.
• Eventual consistency replicas converge over time.
• CAP theorem states that it is impossible for a distributed system to simultaneously provide consistency, availability, and partition tolerance.
• Choosing the right consistency model depends on the application requirements.

Fault Tolerance

• Fault tolerance allows the system to continue operating despite site failures.
• Replication enhances fault tolerance by providing redundant copies of data.
• Failure detection mechanisms identify failed sites.
• Recovery protocols restore the system to a consistent state after a failure.

Advantages of Distributed Databases

• Scalability is achieved by easily adding more nodes to increase capacity.
• Availability is achieved if the system remains operational even if some nodes fail.
• Performance is increased because data locality reduces network traffic.
• Autonomy allows local control over data.

Disadvantages of Distributed Databases

• Complexity makes it more difficult to design and manage.
• Security creates increased security risks due to distributed nature.
• Control and maintaining data consistency is challenging.
• Overhead includes additional requirements for communication and coordination.

Use Cases

• Financial institutions use distributed databases for transaction processing.
• E-commerce platforms use them for managing product catalogs and customer data.
• Social media networks use distributed databases for storing user profiles and content.
• Cloud computing providers use distributed databases for scalable data storage.