A group project for a distributed systems course in uni. A simple music streaming service implemented in a client-broker-server model. This distributed application consists of 4 kinds of nodes, the Publishers(or servers), the Brokers,the Zookeeper, and the Consumers(or clients). The Publishers, the Zookeeper and the Brokers run Java apps on many computers (or a single computer but using different ports), while the Consumers run an Android application. The nodes communicate via network using Java sockets.
Each node has a different purpose. Briefly, the Publishers store mp3 file data and forward them to the Brokers, each time a Consumer requests some song data. The Brokers, then, forward the requested data to the Consumers. Load balancing is achieved via the use of many Brokers . Each Broker is answering requests for a range of songs ,ex. a range of song name initials. A Broker answers Consumer requests for songs starting with A-C, another one for songs starting with D-F, etc. The distribution of song names to the Brokers is achieved with SHA-1 hashing, in a way similar to the CHORD peer-to-peer protocol.We used 3 brokers for this assignment. Parallel programming using threads is also used, so that many Consumers can send requests for the same songs simultaneously.
This node is used to synchronise the Brokers, so that all Brokers know which song requests to answer. The Zookeeper could possibly do more things as in the Apache Kafka stream-processing software platform, but we kept it simple for this project.
The Consumers are running a simple Android application that can run both online or offline. If online, the app is initially requesting song information from the brokers and showing the song names in a scrollable list. The user can select an artist and a song, and the application instantly sends a request for the song to the right Broker. Then, the user can listen to the music with a simple player. Clients can also download songs. If offline, the user can listen to downloaded songs only.
The song datasets are copyright-free and can be downloaded here
We used the mp3agic library to read mp3 file tags.
Download the datasets from the link above, and place them in the Event Delivery System folder, so that Event Delivery System will eventually have J-K and 0-J folders directly in it. Each Publisher has its own songs located in one of the folders stated above, and the connection between a Publisher and a folder is done in code. Therefore, in case a folder is renamed, Publisher main classes must also be updated.
In the Event Delivery System/src/../Node.java file there are lists with the addresses of Publishers, Brokers, Zookeeper. We change the IP addresses manually. The same data must be defined in the App Android/app/.../Node.java.
We execute the main apps in the following order:
-> first Zookeeper.java.
-> then Broker1.java, Broker2.java, Broker3.java, in any order. Each of these files contains a main method, one for each Broker.
-> then Publisher1.java, Publisher2.java (main for Publishers), in any order.
-> finally, we run the Android application on a mobile phone or Android Studio VM, for the Consumer.
Themelina Kouzoumpasi
Lydia Athanasiou
Konstantina Souvatzidaki