Distributed Computing

This project's goal is to implement a Java RMI distributed computing with load balancing. Many servers can connect to our load balancer and are available to receive calculous. A server may or may not give the right answer and each server has it own capacity (beware of overload !)

In order to handle server overloading, the load balancer adjuts the number of calculatons sent to every server.

In order to handle malicious servers, the load balancer has 2 different modes SAFE & UNSAFE:

• SAFE => Server results are trustful and the repartitor only merges the results.
• UNSAFE => Server aren't trustful and the repartitor always verificates the server's results with cross-validation

Compile this project

• Go to v0 folder.
• compile sources : ant

Launch servers

• launch rmiregistry on port 5010 from ./bin :

# rmiregistry PORT &
rmiregistry 5010 &

• launch server :

# ./calculousServer IP PORT CONFIDENCE CAPACITY
./calculousServer 127.0.0.1 5010 0 5

Launch load Balancer

• Server Ips and server ports configuration should be in config/servers.config:

127.0.0.1 5010
127.0.0.1 5011
190.0.0.1 5010

• Launch Load Balancer :

# ./repartitor [-S] : calculous are computed a 2nd time (cross-validation)
./repartitor -S

Distribute computing

Once the load repartitor is launched, we have 2 options compute FILE & exit:

• compute FILE => it parses the input file and distributes computing with the differents available servers, then prints the result.
• EXIT => it kills the load balancer.

operation files are available as operations-XXXX where XXXS is the file result

French report is available in V0 folder : Rapport-TP2-INF4410.pdf

Author

• Jérémy Wimsingues
• Robin Royer