Building Data Processing Systems with FPGAs (Reading Group)

Seminar offered in UPM Master Universitario en Software y Sistemas, Spring Semester 2019

Instructor

Zsolt Istvan, IMDEA Software Institute

E-mail: zsolt.istvan [at] imdea.org

Outline

This 1 ECTS seminar is aimed at students who are familiar from their earlier studies with FPGAs and would like to learn about use-cases beyond the traditional ones of embedded computing and signal processing. To this end, we will explore together the novel trends in the area of hardware accelerators for databases and Cloud-scale data processing systems. Additionally, we will discuss how the inclusion of specialized hardware in complex software systems changes the ways we think of software engineering and development.

We will cover topics related to database, machine learning and distributed system acceleration using FPGAs. We will discuss their benefits, drawbacks, and potential research challenges together in a reading group.

The final selection of papers will depend on the number and interests of attendees but below are some examples:

• FPGA: What's in it for a Database? R. Mueller et al.
• Azure Accelerated Networking: SmartNICs in the Public Cloud. D. Firestone et al.
• Caribou: Intelligent Distributed Storage. Z. Istvan et al.

Teaching Methodology

Prerequisites

• Basic understanding of FPGA programming (at least one introductory course in your Bachelor's!)
• Basic understanding of computer architecture
• Basic understanding of databases

Structure and Location

The course will take place at IMDEA Software Institute, in the Montegancedo Campus, room 315.

Meetings:

• 13.02 -- Introduction [Slides][Reading]
• 27.02 -- Student A: BlueCache   Student B: Caribou
• 06.03 -- Student A: Skylines   Student B: Data Processing
• 13.03 -- Student A: FINN   Student B: Regex Reconf.
• 20.03 -- TBA

Preparation at home: approx. 5 x 3h

Grading

• Presence at meetings: 40%
• Paper presentation: 40%
• Participation in discussions: 20%