CS435 - Distributed Systems<br>
with Dr. Basit Qureshi<br>
<br>
Tutorial 7<br>
_____________<br>

Read through this tutorial on Synchronization Algorithms.

<br><br>
_____________<br><br>
1. Define Distributed Mutual Exclusion. What are the goals to achieve in any solution for this issue?[S 5]<br><br>
2. Explain how a FIFO queue can help with a centralized algorithm. Why it causes a bottleneck? [S 8-13]<br><br>
3. What is the "cost" of Lamport Mutual Exclusion algorithm in terms of communication? [S 24]<br><br>
4. In a ring based algorithm, explain what happens when a node fails? Using the following figure, give a step by step illustration with messages. Assume node 24 (leader) crashed. Show communication between 15 and 24. [S37-45]<br><br>
<img src="RING.jpg" ><br><br>
5. In bully algorithm, what gives if bully is not responsive?<br><br>
6. Explain why maintaining a quorum is essential in concensus algorithms? [S 47-54]<br><br>
7. Explain why managing network partitions is such a pain? [S 46]<br><br>
8. Read the <a href="https://dl.acm.org/doi/10.1145/3380787.3393681" target="_blank">Paxos vs RAFT paper</a> [<a href="Paxos-RAFT.pdf" target="_blank">pdf</a>] Answer the following:<br><br>
<li>How does each algorithm handle leader election, and what are the implications of their respective approaches?</li>
<li>How do Paxos and Raft ensure consistency and data integrity in distributed systems, and what are the differences in their approaches to maintaining consistency?</li>
<li>Discuss the scalability characteristics of Paxos and Raft, including their ability to handle large-scale distributed systems and increasing numbers of nodes?</li>
<li>What are the recovery mechanisms employed by Paxos and Raft in the event of network partitions, and how do they ensure system resilience and consistency?</li>
<br><br><br><br><br><br><br><br><br><br>