1. Implementation of Ricart Agrawala Algorithm
By: Rahul Khadse
Computer Science Department
Kent State University

2. INTRODUCTION My Project implements Ricart Agrawala Algorithm .
Objective :-
Implement the Algorithm using C++.
Modify the Algorithm
Study the performance of the Algorithm.
To understand the behaviour of the algorithm in different cases.

3. Short Review of Algorithm
Ricart Agrawala is optimization of Lamport's Algorithm that dispenses with Release messages by cleverly merging  them with Reply messages
Reduced message complexity :  2(N-1)
Lamport Algorithm
Ricart Agrawala
Request (N-1)
Reply(N-1)
Release(N-1)
Request(N-1)

4. Ricart Agrawala Experimentation Setup :
Original Algorithm :
Vary the no of nodes in system (size).
Vary the No of nodes accessing CS (Load)
Modification of Ricart Agrawala Algorithm
The Algorithm tends to favor lower numbered nodes slightly , owing to tie breaking rule
This Favoritism can be reduced by incrementing the SQN by higher no for lower numbered nodes

5. Ricart Agrawala Experimentation Setup :continued
Process
Increment P0 20 P5 1 P1 P6 P2 P7 P3 P8 P4 P9
My expected result was : After modification no of CS accesses made by low numbered processes should increase than high numbered processes.

6. Results (Orginal Algorithm)

7. Results (Orginal Algorithm) contd..
Message Complexity doesnot change in case of High Load and Low Load .
Message Complexity changes with no of processes in the system.

8. Results (Modified Algorithm)
When total no of CS accesses are 200 .

9. Results (Modified Algorithm) continued..

10. Results (Modified Algorithm) continued..

11. Future Research
Making Ricart Agrawala to work in practical networks  :-
Insertion of new nodes
Removal of Nodes
Failure of Nodes
Hierarchical Ricart Agrawala can be implemented in case there are multiple sites and several no of processes working in each site.

12. Conclusion
Message complexity of Ricart Agrawala is not depend on high load or low load but depends on the number of processes in the system. Thus Ricart Agrawala gives constant performance even in case of high load.
In modified Ricart Agrawala irrespective of sequence number increment by higher number for low numbered (Higher Priority Process) , the number of CS accesses made by high priority and low priority processes are same.

13. Code Design and Difficulties
Simulator :-
Creating the instance of Processes.
Running the implementation of Ricart Agrawala algorithm
Function MUTEX which invokes mutual exclusion for particular node
Function RECEIVE for Receiving Request or Reply.
Difficulty :-
For designing function MUTEX I need to access the data of all Processes . But C++ don't allow to pass a complete block of message by value as a parameter to a function
Hence I used vector and passed it as a parameter by reference instead of by value .

14. Thank You
Questions ?