Sistem Terdistribusi: Transaction

Summary

This lecture on distributed systems focuses on distributed transactions, exploring concepts like atomicity, concurrency control, and the two-phase commit protocol. It covers the challenges of transaction coordination across multiple servers, using protocols like the 2-phase and 3-phase commit to ensure consistency. The session also addresses deadlock issues, recovery mechanisms, and the role of coordinators in managing transactions. Additionally, the importance of synchronization between servers to maintain data integrity is emphasized, with examples from banking systems and ATM transactions.

Takeaways

• 😀 Distributed transactions involve multiple servers working together to process a single transaction across a network.
• 😀 Atomicity ensures that either all parts of a distributed transaction are completed successfully or none at all. If one part fails, the transaction is rolled back.
• 😀 Concurrency control is essential to manage multiple transactions happening simultaneously without causing data inconsistency or conflict.
• 😀 Two-Phase Commit (2PC) is a protocol used to ensure all servers either commit or abort the transaction, but it can experience blocking issues in case of network failures.
• 😀 Three-Phase Commit (3PC) is an extension of 2PC designed to prevent blocking during network partitioning by using an additional phase for greater reliability.
• 😀 Deadlock can occur when transactions wait indefinitely for each other to complete, causing the system to halt. Managing deadlock is a key concern in distributed systems.
• 😀 Recovery mechanisms, such as transaction logging, help restore data to a consistent state after a system crash or failure.
• 😀 Distributed systems require a **coordinator** to manage the transaction process, ensuring all involved servers reach a consistent decision to commit or abort.
• 😀 Concurrency control can involve mechanisms like **locking** and **timestamping** to prevent conflicts when multiple transactions access the same data.
• 😀 A distributed transaction coordinator works like a **foreman**, managing the transaction process, determining when to commit or abort, and ensuring all servers in the transaction are consistent.
• 😀 When a system crash occurs, recovery mechanisms like **rollback** ensure that the system can revert to the last committed state of the transaction, preventing data loss or inconsistency.

Q & A

What is the main topic discussed in this session?

The session discusses distributed transactions, focusing on concepts such as transaction coordination, protocols, concurrency control, and recovery mechanisms in distributed systems.

What is a distributed transaction?

A distributed transaction involves multiple servers working together to process a request from a client, ensuring that the transaction is either fully completed or fully rolled back across all involved systems.

What is the role of a transaction coordinator in a distributed system?

The transaction coordinator acts as a manager that ensures all participating servers agree to either commit or abort the transaction. The coordinator is responsible for synchronizing the process across multiple servers.

What is the concept of 'Atomic Commit Protocol'?

The Atomic Commit Protocol ensures that a transaction is either fully committed or fully aborted. It guarantees that all systems in the distributed transaction agree on the transaction's outcome.

How does the 'two-phase commit' protocol work?

In the two-phase commit protocol, the first phase involves the coordinator sending a 'prepare to commit' request to all participants. The participants respond with either 'yes' or 'no'. In the second phase, if all participants agree, the coordinator sends a 'commit' message; otherwise, it sends an 'abort' message.

What is the difference between the 'two-phase commit' and 'three-phase commit' protocols?

The three-phase commit protocol addresses some of the limitations of the two-phase commit protocol, such as the possibility of blocking due to network issues or participants failing to respond. It introduces an additional phase to reduce the risk of waiting for a response indefinitely.

What are some of the concurrency control mechanisms in distributed transactions?

Concurrency control in distributed transactions can be achieved through methods like locking (ensuring exclusive access to data), timestamping (ordering transactions based on time), and optimistic concurrency control (allowing transactions to execute and checking for conflicts later).

What is a potential issue with locking in distributed transactions?

Locking can lead to problems like deadlock, where transactions wait indefinitely for each other to release resources. This can occur when two or more transactions hold locks on resources that the others need.

What is deadlock in the context of distributed transactions?

Deadlock occurs when two or more transactions are each waiting for the other to release a resource, causing a situation where neither transaction can proceed. This often requires intervention to resolve.

How does recovery work in distributed transactions?

Recovery in distributed transactions involves using a recovery manager to restore the system to a consistent state after a failure. This can include rolling back unfinished transactions or restoring data from a stable storage medium based on transaction logs.