MPCS 52040 Distributed Systems (Spring 2016)

Section 1
Location Young 302
Meeting Times Thursday 5:30pm - 8:30pm
Fulfills Core Systems Elective


Course Description
This class teaches the theory and practice of how to design a distributed system. 

Cloud computing, drive services, online collaborative working environment, massively multiplayer online gaming, airline reservation systems, e-commerce are examples of distributed systems. Because of a continuous growing number of independent and diverse connected devices, the need of managing heterogeneous architecture became prominent. We will learn how to create a reliable system being easy to use but handling the complexity of having all these machines work collectively.
The objective of this class is to learn distributed system by studying:
  • Communication mechanisms
  • Synchronization
  • Scalability
  • Consistency / Replication
  • Fault tolerance
Course Organization
This course meets once a week for lectures and once a week for an optional TA session. Even if not recorded, attendance in class is strongly encouraged. Most of the coursework revolves around 1 programming assignment, as the primary focus will be for students to code. There will be one presentation (Week 5) and one Final Exam (Week 10). The course calendar, including the contents of each lecture is shown at the end of this page.

On Week 11, volunteers will present a topic on distributed system (the grade will provide extra points to your homework assignment).
The final grade will be based on the following:
  • Written assignments: 15%
  • Programming assignments: 50%
  • Midterm Exam (project presentation): 10%
  • Final Exam: 25%
Written Assignments
We will assign simple written problems at the end of each lecture that will be due the following week. The goal of these assignments is to make you understand the course better. Some content can be hard to appreciate in class, and we encourage you to work them at home the same day. They will constitute 15% of the course grade.
Programming Assignment
We will assign a programming assignment over the course of the term. You can choose to work on the programming assignment alone, however, it is strongly recommended to work by group of 2 people. You will need to inform us of your choice by Week 2.
The goal of this assignment will be to make you create your own distributed system. It will constitute 50% of the course grade.
In this project, you will design, implement and test a worldwide trading system platform. You will need to present your work during the mid-term by Week 5. During this project, you will create a system capable to handle trading across countries.
Late Submission
Prorated points will be applied according to the number of late days.
1 day late: 75% of the grade will be considered.
2 days late: 50% of the grade will be considered.
3 days late: 25% of the grade will be considered.
After 3 days, no assignment will be accepted.
We will be giving a final exam at Week 10. The final exam is worth 25% of the grade.
The final exam will be a closed-book written exam on the whole content of the class.
The questions of this exam will be on the content of the lecture and the programming assignment.

Distributed Systems: Principle and Paradigms, Second Edition, Andrew Tanenbaum, Maarteen Van Steen

Course Contents

Lecture 1: 
Introduction to Distributed Systems.
Definition, Example, Type of Distributed Systems (Computing, Informative, Booking)
Lecture 2:
Processes / Communication
Operating System - Tasks / Threads / Synchronization
Communication: layered protocols
Lecture 3:
Clock synchronization, Logical clocks, mutual exclusion, election algorithm
Go Language
Lecture 4:
Networking / Internet / QOS / Latency / Routing
Remote Procedure Call
Lecture 5:
"Mid-term" - Project suggestion from students
Description of the application / Overall structure / Distributed System / Plan for testing the system
Lecture 6:
Naming / Mapping
Lecture 7:
Fault Tolerance / Consistency
Byzantine fault tolerance
Lecture 8:
Distributed File System
Architecture / Communication / Naming / Synchronization 
Lecture 9:
Data-Intensive Computing - MapReduce
Lecture 10:
Final exam
Lecture 11:
Advance distributed system presentation
- Support multiple agents
- Support multiple client/server nodes
- Must be reliable/robust (crash recovery)
- Use of some algorithms covered in the class (distributed synchronization, replication, fault tolerance, recovery, security)

Example: shared document editing, simulated life game, multi-player game, low-latency event system (trading system)
The project should be proposed by students and analyzed during the lecture 3.

Course Prerequisites

Core Programming

Other Prerequisites

Overlapping Classes

This class is scheduled at a time that conflicts with these other classes:

  • MPCS 55001-1 -- Algorithms