Welcome

Yael Cloud Platform

(tssDas)

by Tsur, Schneider & Shani

[Topic List]

This is our final year project in MTA college made by computer science students. The system will allow the distribution of different pieces of input data to multiple instances of an algorithm written in java (and confirming to our basic requirements) to be run on a cloud.

This project is dedicated to our dear friend [Yael Hadar] who is, and will be sorely missed.

To view the project slide you may go here.

Project Members:

1. Doron Tsur.
2. Yaron Schneider.   
3. Amir Shani.

Here is the general out line of the project

Basic Requirements.

Same algorithm - each machine runs the same algorithm on a different part of the data.
Supply from user - the system user will develop and test the data multiplexer, de-multiplexer and the actual algorithm in a manner that will fit our system.

Main components - Milestone 1

Lead Server - The server is responsible for maintaining communication with all worker servers, providing them with the algorithm and all relevant data for their partial calculations. It will also hold the result of a the full algorithm run. Intermediate results will be kept in a DB on the lead server.
Worker Server - The worker server's main job is to communicate with the server ,receiving and running the algorithm on slices of data sent to it by the lead server.

Main work flow - (main UC)
After deploying Yael the user could develop a java based algorithm and run it on the platform's resources. The lead server will address the worker stations providing them with the algorithm and data slices to analyze. The worker server will execute the algorithm, send back the result slices and address the server for more data slices. This will continue until algorithm is completed or aborted.
for a full spec see: [Yael Cloud platform - Spec]

Backup - Milestone 2.(not in scope of this project)

Backup server - the Lead Server (LS) will provide backup data for restoration in case it crashes. The backup server will allow to restore a LS to it's last saved snapshot.

Work Flow- Backup
Server crashes for some unknown reason. an operator will ask it to restore from backup and continue. The server will address all active Worker Stations (WS) and ask for last results or prompt them the next task. The main work flow will continue.

Take Lead - Milestone 3.(not in scope of this project)

Sleeping server - each WS will be able to become a LS in case LS crashes.It will address the backup and restore via network restore.

Work Flow - Auto restore
LS crashes, stations will try and connect to it and fail. They will connect to a backup server and ask for new LS. If there is no new LS they will become one and all stations that will address the BS afterwords will be forwarded to the new LS. New LS will start work flow backup