2018-2019 / INFO8002-1

Large-scale data systems

Duration

25h Th, 10h Pr, 45h Proj.

Number of credits

 Master in data science (120 ECTS)5 crédits 
 Master of science in computer science and engineering (120 ECTS)5 crédits 
 Master in data science and engineering (120 ECTS)5 crédits 
 Master in computer science (120 ECTS)5 crédits 

Lecturer

Gilles Louppe

Language(s) of instruction

English language

Organisation and examination

Teaching in the first semester, review in January

Schedule

Schedule online

Units courses prerequisite and corequisite

Prerequisite or corequisite units are presented within each program

Learning unit contents

In the modern data landscape, large-scale data systems have become a critical component of the data science analysis pipeline. They are of primary importance for the reliable storage, but also for the analysis, of the increasingly larger volumes of data encountered in web-based applications, cloud computing centers or in networks of connected objects. 
However, large-scale data systems remain notoriously difficult to build because they need to scale to hundreds or thousands of machines, they must be tolerant to crashes, they have to cope with concurrent execution and they need to ensure consistency of the data they store. 
In this context, the course will cover concepts in a bottom-up fashion. We will first cover the foundational abstractions that are the core of distributed systems, including basic abstractions and system assumptions, reliable broadcast, shared memory and consensus. We will then study data computing systems that are built on top of those components, including MapReduce and computational graph systems (Spark). Similarly, we will study distributed storage systems, including distributed file systems, distributed key-value stores and block chains. 
Topics to be covered:

  • Data deluge
  • Basic distributed abstractions
  • Reliable broadcast
  • Shared memory
  • Consensus
  • Blockchain
  • Distributed hash tables
  • Cloud computing
  • Distributed file systems

Learning outcomes of the learning unit

At the end of the course, the student will have understood the core building blocks of reliable distributed systems. He/she will also have acquainted with industrial data systems and their inner workings. Finally, he/she will have developed a critical thinking regarding the benefits and limitations of these systems in the context of data science needs. 

Prerequisite knowledge and skills

Programming experience. Basic knowledge in computer networks.

Planned learning activities and teaching methods

  • Theoretical lectures
  • Exercise sessions and programming tutorials
  • Reading assignment
  • Programming project (e.g., implement a simple data system).

Mode of delivery (face-to-face ; distance-learning)

Lectures will taught face-to-face. Projects will be carried out remotely.

Recommended or required readings

Slides will be made publicly available on GitHub during the semester.
Part of the course will be based on "Introduction to Reliable and Secure Distributed Programming", Christian Cachin, Rachid Guerraoui, Luis Rodrigues, Springer. This book is recommended.

Assessment methods and criteria

  • Exam (60%)
  • Reading assignment (10%)
  • Programming project (30%)
The reading assingment and the programming project are mandatory for presenting the exam. 

Work placement(s)

Organizational remarks

The website for the course is https://github.com/glouppe/info8002-large-scale-database-systems

Contacts

  • Teacher: Prof. Gilles Louppe (g.louppe@uliege.be)
  • Assistant: Joeri Hermans (joeri.hermans@uliege.be)