13
QUESTIONS – Distributed Computing Systems Prof. Ananthanarayana V.S. Dept. Of Information Technology N.I.T.K., Surathkal

QUESTIONS – Distributed Computing Systems · Questions ‐Fundamentals. 1. Discuss five fundamental issues in distributed system 2. What are different transparencies which can be

  • Upload
    dangdan

  • View
    221

  • Download
    6

Embed Size (px)

Citation preview

Page 1: QUESTIONS – Distributed Computing Systems · Questions ‐Fundamentals. 1. Discuss five fundamental issues in distributed system 2. What are different transparencies which can be

QUESTIONS – Distributed Computing Systems

Prof. Ananthanarayana V.S.

Dept. Of Information Technology

N.I.T.K., Surathkal

Page 2: QUESTIONS – Distributed Computing Systems · Questions ‐Fundamentals. 1. Discuss five fundamental issues in distributed system 2. What are different transparencies which can be

Questions ‐ Fundamentals

1. Discuss five fundamental issues in distributed system2. What are different transparencies which can be observed in distributed system? 

List basic transparencies need to be supported by the distributed system.3. Give main two technical differences between Network OS and Distributed OS.4. Consider a distributed environment with four systems, A,B,C and D. Name the 

type of transparencies required in each of the following situations:a. Data available at all four systems and user want to modify the data at D.b. Printer connected to A is disconnected and connected to B. User wants to access that 

printer.c. Breakdown  of system, D.d. User want to access the software X without knowing its whereabouts.

5. Consider that railway reservation system is implemented using distributed environment. List out the possible types of transparencies need to be incorporated in this system. Justify your answer.

6. List any two situation which clearly shows the need for distributed coordination. How this can be achieved?

7. Explain two types of resource management in distributed system.8. Give characteristics and goals of centralized OS, network OS, distributed OS and 

cooperative autonomous system.

Page 3: QUESTIONS – Distributed Computing Systems · Questions ‐Fundamentals. 1. Discuss five fundamental issues in distributed system 2. What are different transparencies which can be

Questions – Distributed Coordination

1. What are causally related events?2. Show that using single value for logical clock C, it is not possible to 

ensure that if C(a) < C(b), then a→b, where → denoted happen‐before relation among the events in a distributed system.

3. Give implementation rules of Lamport’s logical clock and vector clock. In which type of applications vector clocks are more appropriate?

4. Why global clock is important in a distributed system? Why this is an issue? How this can be realized?

5. Consider a bank database which is fully replicated. Give an algorithm/protocol for ordering of transactions in above situations.

6. Differentiate implementation rules of Vector clock and Lamport’sclock.

7. Prove that vector clock condition is strong. What is the significance of this condition?

Page 4: QUESTIONS – Distributed Computing Systems · Questions ‐Fundamentals. 1. Discuss five fundamental issues in distributed system 2. What are different transparencies which can be

Questions – Ordering of Messages

1. Trace SES protocol to ensure the ordering of messages in above scenario.

2. Discuss BSS protocol for causal ordering of messages. In what way this algorithm is different from SES protocol?

Space Time

e11 e12

e21 e22

e31 e32

Page 5: QUESTIONS – Distributed Computing Systems · Questions ‐Fundamentals. 1. Discuss five fundamental issues in distributed system 2. What are different transparencies which can be

Questions – Global State Detection1. Why global state detection is an issue in distributed system?2. Give the consistent global state requirements in DCS. When the global 

state is said to be strongly consistent?3. Let ‘C’ be the channel from node S1 to S2. Show that a consistent global 

state must always satisfy n≥m, where ‘n’ is the number of messages sent by S1 along ‘C’ before S1 state is recorded and ‘m’ is the number of messages received by S2 along ‘C’ before S2  state is recorded.

4. Consider a DCS with set of sites, S={S1,S2,S3}. With respect to real time, t, the global state G={LS1, LS2,LS3}  (where LSi is local state of site Si) is recorded.Let LS1={recv(e21, e11), send(e12, e31), recv(e22, e13)}Let LS2={send(e21, e11), send(e22, e13), recv(e14, e23)}Let LS3={recv(e12, e31), send(e32, e15), recv(e16, e33)}where eij is jth event at ith site; send/recv(x,y) – send/receive message from x to y. Comment with justification on global state G recorded.

5. Differentiate:Transit‐less global state from Consistent global state

Page 6: QUESTIONS – Distributed Computing Systems · Questions ‐Fundamentals. 1. Discuss five fundamental issues in distributed system 2. What are different transparencies which can be

Questions – Distributed Process Synchronization

1. In Raymond’s tree based DME algorithm, what happens to the message overload when arrival rate of critical section requests increases at each node?

2. How deadlock situation is handled in Meakawa’s DME algorithm?3. Compare the performance of Ricart‐Agrawala’s token based DME algorithm with 

Ricart‐Agrawala’s permission based algorithm4. If communication channel is NON‐FIFO, does 

a. Lamport’s DME algorithm ensures mutual exclusion condition?b. Ricart‐Agrawala’s permission based DME algorithm ensures mutual exclusion condition?

5. “Token based DME algorithms are less fault tolerant than permission based algorithms” Comment on this.

6. Which one of the following algorithm is fair with respect to Lamport’s clock? Why?

1. Meakawa’s DME algorithm2. Ricart‐Agrawala’s DME algorithm3. Raymond’s tree based DME algorithm

7. Differentiate:1. Fair DME algorithm from starvation free DME algorithm2. Framework for DME algorithm from its centralized counterpart.3. Permission based DME algorithm from token based DME algorithm4. DME algorithm performance measure: Response time from Waiting  time.

Page 7: QUESTIONS – Distributed Computing Systems · Questions ‐Fundamentals. 1. Discuss five fundamental issues in distributed system 2. What are different transparencies which can be

8. Give pseudo‐code for 1. Lamport’s DME algorithm2. Ricart Agrawala’s permission and token based DME algorithm3. Meakawa’s DME algorithm4. Raymond’s tree based DME algorithm

9. Compare the performance measures between following algorithm:1. Lamport’s DME algorithm2. Ricart Agrawala’s permission and token based DME algorithm3. Meakawa’s DME algorithm4. Raymond’s tree based DME algorithm

10. What is the necessity of INQUIRE and RELINQUISH message in Meakawa’s DME algorithm?

11. What is the problem if the topology considered for Raymond’s algorithm has a cycle?

12. How you can convert the following algorithm into greedy algorithms?1. Ricart‐Agrawala’s token based DME algorithm and 2. Raymond’s tree based algorithm

Questions – Distributed Process Synchronization

Page 8: QUESTIONS – Distributed Computing Systems · Questions ‐Fundamentals. 1. Discuss five fundamental issues in distributed system 2. What are different transparencies which can be

Questions – Inter Process Communication

1. What are the various message orderings that are used in group communication?

2. What are different failure handling semantics in RPC?3. List various semantics of send primitives. Give appropriate application 

for each of the semantics4. What is atomic broadcast protocol? Give general overview of its 

implementation. 5. Differentiate system call and local procedure call.6. Discuss the system calls which are used to provide the communication 

1. Between related processes2. Between the processes within the file system3. Between any arbitrary processes

7. How parameter and result passing is handled in RPC?8. Give block schematic to explain how a single process can communicate 

with two of its child processes – using pipes? What are the limitations of pipes?

9. Consider a client‐server application. Which semantics of send is preferred? Why?

Page 9: QUESTIONS – Distributed Computing Systems · Questions ‐Fundamentals. 1. Discuss five fundamental issues in distributed system 2. What are different transparencies which can be

10. Give block schematic which shows the various steps in RPC handling.

11. It is required to update a bank account in DCS in order. Which message ordering is appropriate? Why? 

12. List the requirements of broadcast protocol.13. What is indirect communication? Give an application where this 

type of communication is appropriate.14. What are the roles of binding register in RPC? In order to load 

balance the service providers, what is the requirements of binding register

15. What are pros and cons of different types of binding?16. What do you mean by non‐blocking RPC? Give an example.17. Differentiate unanimity and uniformity requirements of broadcast 

protocol. 

Questions – Inter Process Communication

Page 10: QUESTIONS – Distributed Computing Systems · Questions ‐Fundamentals. 1. Discuss five fundamental issues in distributed system 2. What are different transparencies which can be

1. Differentiate reusable and consumable resources with example.

2. What are different resource request models?3. Which types of deadlocks are handled by wait for graph 

(WFG)?4. Which types of deadlocks are not handled by WFG? How 

this can be handled?5. Show that how wait‐die / wound‐wait protocol ensures 

prevention of deadlocks?6. How wait‐die / wound‐wait protocol can be extended to 

handle deadlock prevention in more than one process waiting scenarios?

7. What do you mean by false deadlocks in distributed environment? How this can be detected?

Questions – Distributed Deadlocks

Page 11: QUESTIONS – Distributed Computing Systems · Questions ‐Fundamentals. 1. Discuss five fundamental issues in distributed system 2. What are different transparencies which can be

8. Consider Chand‐Misra‐Haas distributed deadlock detection and recovery algorithm. How an unique victim is identified if resource request model is (a) single unit    (b) AND  resource request model?

9. Which is the deadlock prevention method where there is no resource pre‐emption? Why?

10. Explain the situation where in a cycle in WFG does not implies the deadlock. How the deadlock can be detected in such  situations?

11. Consider an OR resource request model and WFG, G = {V,E}, where V = {P1, P2, P3} and E = {<P1, P2>, <P2, P3>, <P3, P1>}. Is the system in deadlock state? Justify.

Questions – Distributed Deadlocks

Page 12: QUESTIONS – Distributed Computing Systems · Questions ‐Fundamentals. 1. Discuss five fundamental issues in distributed system 2. What are different transparencies which can be

1. Differentiate CPU scheduling from Job scheduling. 2. Differentiate Job scheduling with load balancing.3. What are the objectives of load balancing? 4. Why and what knowledge about the node is important in 

load balancing? What are the issues if knowledge estimated is not correct?

5. What is the advantage of sender initiated algorithm over receiver initiated algorithm?

6. Compare transfer and location policy between sender initiated and receiver initiated algorithm

7. What are the characteristics of above average algorithm?8. How the system load is estimated in above average 

algorithm?

Questions – Load Balancing 

Page 13: QUESTIONS – Distributed Computing Systems · Questions ‐Fundamentals. 1. Discuss five fundamental issues in distributed system 2. What are different transparencies which can be

9. What are the drawbacks of above average algorithm when compared with stable symmetrically initiated algorithm?

10. What is “learning process” in stable symmetrically initiated algorithm?

11. Explain location policy of stable symmetrically initiated algorithm.

12. What is the advantage of stable sender initiated algorithm over stable symmetrically initiated algorithm?

13. What is the use of  “status vector” in stable sender initiated algorithm?

14. What are the performance parameters for  a load balancing algorithm?

Questions – Load Balancing