Hasso-Plattner-Institut
Prof. Dr. Tilmann Rabl
  
 

Data Processing on Modern Hardware

Instructors

Prof. Dr. Tilmann RablPedro SilvaLawrence BensonIlin TolovskiWang Yue

Contents

In this project seminar, we will discuss data processing techniques on modern hardware. Specifically, we will look at the characteristics of modern processors (GPU, FPGA), memory (NVRAM), and network (RDMA) and research, how data processing can be done most efficiently on these devices. To this end, we will survey current trends in modern hardware, read and present research papers on data processing on modern hardware, identify a small research project and implement and evaluate a prototype for data processing on modern hardware.

 

Announcements

  • Course management will be done using the Algorithm Engineering Moodle.
  • The moodle page for the course is now open for students. You can enroll here.
    • Once you have enrolled in Moodle, please indicate in the "I want to participate" question whether you want to participate or not. This way we get an estimate of how many people are interessted.
  • There is an intro video available on Moodle.
  • Non-HPI participants: please send us an email to get access to the Moodle.
  • Due to the corona situation, we will probably start the course fully online on April 21. More details to follow.

 

Basic requirements for this course are:

  • Good programming skills (ideally C/C++)
  • Interest in modern hardware
  • Interest and previous knowledge in database systems (e.g., DBS II, BDS, DDM)

 

Schedule

The seminar will take place Tuesdays at 15:15 in F-1.04, Building F at Campus II or in Zoom. Even in the fully online setting, we will use this slot for presentations and discussions.

 

Structure (more details in Moodle)

Paper Presentation and Write-up

  • Each student must present (at least) one selected paper.
    • The papers are selected by us to give all students an overview of current research and/or foundations of the different hardware topics. 
    • The list is still being curated, once it is final, we will post it in Moodle.
  • Every week, each student must summarize one of the presented papers in a one-pager.

Project

  • Each student has to participate in a group project (2 people) on a "data processing on modern hardware"-related project of their choice.
  • The projects must be based on one of the hardware technologies covered in this course.
  • Each group will give a 5 minute kick-off presentation presenting their project proposal, covering: problem statement, goal, approach.
  • Each group will give a tutorial ("hello world") presentation on how to use their selected hardware.
  • Each group will implement a research-prototype and evaluate it.
  • Each group will give a 20 minute final presentation on the results of their project, covering: recap of proposal, solution, results.

 

Grading

The course mark is broken down as follows:

  • Oral (65% of the grade)
    • Present paper (25%)
    • Present project proposal (5%)
    • Tutorial Presentation (10%)
    • Final presentation (20%)
    • Give feedback and ask critical questions (5%)
  • Written (35% of the grade)
    • Prepare a paper based on your project (8 two-column pages, 30%)
    • Prepare one-pagers for every week (5%)


Course important dates

  • Project proposals: Sunday, May 10th 23:59
  • Proposal presentation: TBD
  • Tutorial presentations: TBD
  • Final presentations: TBD
  • Final report due: Monday, August 31st

 

Literature

Will be announced in class and on Moodle once it is finalized.