DISTRIBUTED COMPUTING VIVA SET
Module 1: Introduction to Distributed Systems
Q1. What are goals of Distributed Systems?
A: Transparency, scalability, fault tolerance, resource sharing.
Q2. What is transparency?
A: Hides system complexity. (Types: access, location, concurrency, failure, replication,
migration)
Q3. Grid vs Cluster computing?
A: Grid = loosely connected. Cluster = tightly coupled.
Q4. What is a NOS?
A: Manages resources in a network. Systems stay independent.
Q5. What is a DOS?
A: Treats all nodes as a single system with global OS.
Scenario: If one node fails in DOS, what happens?
Answer: DOS handles it through fault tolerance and redundancy.
Q6. What is middleware?
A: Software layer for communication in distributed systems.
Q7. Types of middleware models?
A: Object-based, message-oriented, RPC, event-based.
Module 2: Communication
Q1. What is IPC?
A: Inter-Process Communication – allows processes to exchange data.
Q2. RPC vs RMI?
A: RPC = procedural. RMI = object-oriented (Java).
Q3. What are sockets?
A: Endpoints for network communication.
Q4. Types of message communication?
A: Unicast, multicast, broadcast.
Q5. Benefits of message-oriented communication?
A: Loose coupling, reliability, async communication.
Q6. What is stream-oriented communication?
A: Continuous data flow (e.g., video streaming). Uses TCP/RTP.
Scenario: What protocol for video call?
A: Use RTP over UDP (for speed), buffering handles delays.
Q7. What is group communication?
A: One-to-many message delivery. Ensures reliability and ordering.
Module 3: Synchronization
Q1. Why clock sync is needed?
A: To order events correctly across systems.
Q2. Logical vs Physical clocks?
A: Physical = real time; Logical = event ordering.
Q3. What is Lamport’s clock?
A: Timestamp to order events. (Doesn't show causality)
Q4. Vector clock?
A: Captures causality using an array of counters.
Q5. Non-token based Mutual Exclusion?
A: Lamport’s, Ricart-Agrawala.
Q6. Token-based Mutual Exclusion?
A: Suzuki-Kasami, Raymond’s Tree-based.
Q7. Deadlock detection algorithm?
A: Chandy-Misra-Haas.
Scenario: If a token is lost in Suzuki-Kasami?
A: System must regenerate token using recovery mechanism.
Module 4: Resource & Process Management
Q1. What is load balancing?
A: Evenly distributing tasks to optimize performance.
Q2. Process migration vs code migration?
A: Process = running task moves. Code = just logic/code moves.
Q3. Preemptive vs non-preemptive migration?
A: Preemptive = running task. Non-preemptive = idle task.
Scenario: Server overloads, what to do?
A: Migrate processes to less-loaded nodes (load balancing).
Module 5: Replication, Consistency & Fault Tolerance
Q1. What is DSM (Distributed Shared Memory)?
A: Allows memory sharing across nodes.
Q2. What is replication?
A: Creating copies to improve fault tolerance and performance.
Q3. Strong vs Weak consistency?
A: Strong = instant updates. Weak = delayed sync.
Q4. What is process resilience?
A: System continues even if processes fail.
Q5. Types of failures?
A: Crash, omission, timing, Byzantine.
Q6. What is checkpointing?
A: Saving state for recovery after failure.
Scenario: If replica is outdated, how do you ensure consistency?
A: Use client-centric models like Read-Your-Writes.
Module 6: Distributed File System (DFS)
Q1. What is DFS?
A: Shares files across networked nodes.
Q2. What is file caching?
A: Temporary storage to reduce network usage.
Q3. Why replication in DFS?
A: Increases availability and fault tolerance.
Q4. What is NFS?
A: Network File System – allows remote file access.
Q5. Google File System (GFS)?
A: DFS built by Google for large-scale data, highly fault-tolerant.
Q6. MapReduce advantages?
A: Parallelism, fault tolerance, scalability.
Scenario: What if multiple users access the same file?
A: Use locking or consistency model to ensure correctness.
Extra Topics Often Asked
Q: What is Bully Algorithm?
A: Highest-ID process becomes leader after others withdraw.
Q: What is Ring Algorithm?
A: Leader selected via token-passing in ring order.
Q: What is Christian’s algorithm?
A: Synchronizes clock using time server.
Q: What is virtual synchrony?
A: Ensures all nodes see group messages in same order.
Q: What is causal ordering?
A: Messages respect cause-effect sequence.
Tip for Scenario Questions
1. If node fails? → Use fault tolerance, replication, or re-election.
2. If tokens/messages are lost? → Use recovery or re-transmission.
3. If heavy load? → Apply load balancing or process migration.
4. If inconsistent time? → Apply clock sync (Christian/Berkeley).
5. If multiple access to file? → Use consistency or locking.
Short
Full Form Description / Where it's used
Form
Used for data exchange between
IPC Inter Process Communication
processes.
RPC Remote Procedure Call Calls a function remotely as if local.
RMI Remote Method Invocation Java-based RPC for object methods.
NOS Network Operating System Manages computers over a network.
DOS Distributed Operating System Manages multiple computers as one.
Short
Full Form Description / Where it's used
Form
Shared memory abstraction in distributed
DSM Distributed Shared Memory
systems.
DFS Distributed File System Enables file access over a network.
GFS Google File System Google’s scalable file system.
NFS Network File System DFS developed by Sun Microsystems.
QoS Quality of Service Ensures performance in streaming.
TCP Transmission Control Protocol Reliable communication protocol.
UDP User Datagram Protocol Faster but unreliable protocol.
RTP Real-time Transport Protocol Used for media streaming.
MOM Message-Oriented Middleware Middleware like Kafka/RabbitMQ.
Common Object Request Broker
CORBA Object-based middleware.
Architecture
IDL Interface Definition Language Used in CORBA for interface specification.
NTP Network Time Protocol Synchronizes time between systems.
Other One-Liner Terms You May Be Asked
Q: What is a daemon?
A: A background process running continuously (like NFS daemon).
Q: What is a stub in RPC?
A: Acts as a proxy between client and server.
Q: What is marshalling?
A: Packing data before sending over the network.
Q: What is unmarshalling?
A: Unpacking received data into usable form.
Q: What is latency?
A: Time delay between request and response.
Q: What is throughput?
A: Amount of data processed in a given time.
Conceptual Understanding
Q1. Why do we need Distributed Systems?
A: To improve availability, scalability, and fault tolerance.
Q2. Can two nodes have the same logical clock value?
A: Yes, but it doesn't imply causality unless vector clocks are used.
Q3. What happens if the coordinator in Bully algorithm fails?
A: A new election is triggered by the next highest-ID process.
Q4. What is the biggest challenge in Distributed Computing?
A: Coordination and consistency among distributed nodes.
Q5. Why is concurrency an issue in distributed systems?
A: Because processes may access shared resources at the same time.
Scenario-Based Questions
Q6. If one server crashes during file write, what should happen?
A: Use replication and recovery mechanisms to avoid data loss.
Q7. If clocks are not synchronized, what could go wrong?
A: Events may be incorrectly ordered, leading to logical errors.
Q8. What would you do if mutual exclusion fails?
A: Recheck the algorithm’s condition, implement fairness or token passing.
Q9. A node fails in a DFS, how is data preserved?
A: Through file replication and fault-tolerant file access mechanisms.
Difference-Based Viva Questions
Q A
RPC vs RMI RPC is for procedures, RMI is for Java objects.
Q A
Physical Clock vs Logical Clock Physical = real-time, Logical = event-based.
Load Balancing vs Load Sharing Balancing is active, sharing is passive.
Strong vs Eventual Consistency Strong = instant update everywhere, Eventual = delayed.
Client-server vs Peer-to-peer Client-server = centralized; P2P = decentralized.
More Quick Concepts
Q10. What is a stub and skeleton in RPC?
A: Stub = client side; Skeleton = server side proxy.
Q11. What is time drift in clocks?
A: Gradual deviation from actual time due to hardware flaws.
Q12. How do vector clocks solve causality issue?
A: By keeping track of event order across all nodes.
Q13. What is atomicity in distributed systems?
A: Operation either completes fully or not at all.
Q14. What is idempotency in RPC calls?
A: Same call repeated gives same result (safe for retries).
Q15. What’s the use of heartbeat messages?
A: To detect node failures in distributed systems.
Technology-Based or Real-Life
Q16. Name one real-world DFS you use daily.
A: Google Drive / Dropbox / NFS (internally by enterprises).
Q17. What is the role of a master node in GFS?
A: Manages metadata and chunk distribution.
Q18. How does MapReduce handle failure?
A: Failed tasks are re-executed on other nodes.
Q19. Why is token ring better than flooding in elections?
A: Less message overhead, more organized.
Q20. What happens if two processes detect mutual exclusion violation?
A: It means the algorithm failed, could lead to race conditions.