diogofferreira / afternoon-at-races

:racehorse: Concurrent and distributed solutions for a horse racing event :racehorse:

An afternoon at the races

This problem was introduced in the context of the Distributed Systems course (Computer and Telematics Engeneering at University of Aveiro) and consists in designing a concurrent and two distributed solutions of the given problem.

Preamble

Events portray the activities that go by during a typical afternoon at a hippic center, somewhere at the outskirts of Aveiro. There are four main locations: the track where the races take place, the stable wherethe horses rest waiting their turn to enter the competition, the paddock where the horses are paraded before the spectators, and the betting center where the spectators place their bets on the winning horse.

There are three kinds of intervening entities: the pairs horse / jockey participating in the races, the spectators watching the races and placing bets on the horse they hope to win and the broker who accepts the bets, pays the dividends in case of victory and manages in a general manner all the operations that take place.

K races are run during the afternoon, each with N competitors. M spectators are present. The activities are organized as described below

• the broker announces the next race;
• the participating horses are paraded at the paddock by the jockeys;
• the spectators, after observing the horses and thinking about their winning chances, go to the betting center to place their bet;
• the race takes place and one or more horses are declared winners;
• when somebody wins, he or she goes to the betting center to collect the gains.

At the end of the afternoon, the spectators meet at the bar to have a drink and talk about the events that took place.

Each race is composed of a sequence of position increments of the intervening horse / jockey pairs according to the following rules

• the track distance for the race k, with k = 0, 1, ... , K-1, is D_k units of length;
• each horse / jockey C_nk, with n = 0, 1, ... , N-1 and k = 0, 1, ... , K-1 carries out a single positionincrement per iteration by moving randomly 1 to P_nk length units along its path
• the maximum value P_nk is specific of a given horse, because they are not all equal, some being more agile and faster than others;
• the horse / jockey pairs move in parallel paths and may be side by side or overtake one another;
• the winner is the pair horse / jockey C_nk, with n = 0, 1, ... , N-1 and k = 0, 1, ... , K-1, which, after the completion of an iteration and having overtaken the finishing line, has a position with the highest value;
• in case of a draw, all the horse / jockey pairs with the highest position value are declared winners; the dividends to be received are inversely proportional to their number and rounded to unity.

Assuming there are five races, each having four competitors and that the number of spectators is also four.

Active entities life cycles

Broker

Horse / Jockey pair

Spectator

Shared Regions

The following shared memory regions will be accessed by the active entities in mutual exclusion manner:

• General Repository
• Stable
• Control Centre
• Paddock
• Racing Track
• Betting Centre

In the concurrent solution, all of them will be shared memory regions inside the same process addressing space, however, in the distributed solutions, they'll all be running in a different machine (they can also be run on the same, if configured with that purpose).

Solutions

To run all the solutions, Java 8 is needed on all the used machines.

Concurrent Solution

This solution is intended to be run locally and it serves as the base for the distributed versions, since both of them use a Server Replication approach.

To run this solution:

cd afternoon-at-races-concurrent
javac -d out -sourcepath src src/main/AnAfternoonAtTheRaces.java
java -cp out main.AnAfternoonAtTheRaces

Distributed Solution (message passing)

This solution is based on the previous one, since it reutilizes the concurrent code to create multiple proxy agents at the servers to serve multiple clients' requests.

Active entities and clients communicate via message passing through TCP sockets.

It is already provided a script to automatically deploy the application in different machines. Make sure to change the addresses and ports of the remote machines you use to deploy it. Then, just run:

cd afternoon-at-races-distributed
./deploy.sh

Distributed Solution (Remote Method Invocation)

Java already provides a framework to call methods on remote objects through TCP. Therefore, the previous solution was adapted to implement RMI. However, in order to successfully deploy this solution, an HTTP server (Apache, NGINX, etc.) must be configured to serve static files over HTTP (*.class files).

Just like the previous solution, do not forget to modify the deploy script to your remote machines.

cd afternoon-at-races-rmi
./deploy.sh

Diogo Ferreira & Pedro Martins - 2018