Extended Seminar AI for Data Management. The seminar focuses on two topics: learned DBMS components and AI for data engineering tasks. The course starts with a mini lecture series to provide the necessary background for the two practical tasks that follow.
- Dozent*in: Carsten Binnig
- Dozent*in: Jan-Micha Bodensohn
- Dozent*in: Johannes Wehrstein
The goal of this course is to familiarize you with the modern hardware stack present in clouds and datacenters. This is an important goal because, to be able to build and maintain efficient data processing systems (e.g., database management systems, streaming analytics pipelines, machine learning training, etc.), an in-depth understanding of the modern hardware is necessary.
This lecture provides a complete view of the cloud hardware architecture and programming aspects from a software systems aspect and shows how to utilise them best. Throughout the semester we will cover the wide spectrum of hardware which is used today in cloud data centers such as:
- multi-core and multi-socket CPUs
- flash-based storage stacks
- user-space networking
- RDMA and programmable networks
- GPUs and specialized hardware-based accelerators
In addition to the in-depth presentation of how these hardware components are designed and how they work, you will also acquire hands-on experience in programming for them in several coding labs as part of the lecture.
- Dozent*in: Zsolt István
Distributed computation has enabled great scalability in various application areas, e.g., Big Data, Machine Learning, etc., but the correct operation of a large number of compute nodes requires reliable coordination and (meta)data replication. In this seminar we take a deep dive in the underlying algorithms and protocols for coordination and replication, namely consensus algorithms.
The seminar is based on research paper reading at home and discussions/presentations in class. For the first part of the seminar, your instructor will present the seminal works and lead the discussion. In the second part of the seminar, you will take the lead and present the state-of-the-art protocols and discuss their benefits/drawbacks in practice.
This seminar will be useful for students aiming for an industry job or research career in the general area of distributed systems. Familiarity with consensus protocols and understanding of the underlying core concepts is fundamental for building safe, reliable and efficient distributed systems!
- Dozent*in: Zsolt István