Course development

New updated plan with better support for group projects, more variety of topics and revised depth and insights in lectures on Eventual Consistency, CRDTs, Distributed Transactions and Distributed Clock Synchronization (Atomic, Quantum Clocks). Emphasis was put in analysis of existing systems and solutions such as Google Spanner, and trade-offs when it comes to liveness and performance properties.

The course in 2023 featured a mix of more practical projects in addition to theoretical using TLA+ and other model checking tools. In addition, we added a new updated lecture on Distributed Transactions that allowed students to understand the relation between linearizability (Distributed Systems) and strong serializability (Databases). This year we had roughly 8-10 doctoral students also joining the course which we tried as a pilot phase for FID3011 with many successful results such as encouraging doctors to conduct research papers in the field and contribute to the discussions of the course. The paper exam was also introduced in a pilot mode with mixed results as it featured primarily multiple answer questions and not so much reasoning. We therefore decided to have more reasoning-driven assessments in the next iteration

In 2022 the course was actively re-developed to incorporate the needs that we identified in the 2021 version of the course. While the course objectives remained the same this year, most effort was put into making the course address and assess these objectives more reliably. To that end, the most important changes include 1) a remake of all lectures of the course and presentations with richer context, relating all concepts to real problems and systems, 2) A brand new full chapter on scalable data management which includes the important topic of distributed transactions (e.g., two-phase commit), as well as advanced topics in real-time, atomic time, and the future applications of quantum clocks and their properties. This also includes a closer examination of the Google Spanner framework which puts into perspective all advanced topics in the course such as Distributed Consensus, Atomic Clocks, Linearizable Registers, Snapshotting, and Atomic Commitment for Strict Serializability. 3) A wider range of project topics, programming languages of choice, and the choice of student-proposed projects, 4) New labs with rich content on practical consensus, liveness, and leader election as well as a deeper look into TLA+ and model checking. 5) A richer exam that despite being online provided a more accurate assessment through integrated grading criteria, multiple choices, and questions in little time which encouraged “proof of work” and discouraged students to seek answers elsewhere due to the time restrictions. In its current form, the ID2203 is a first-of-a-kind worldwide that adequately covers all these deep topics in a single course.

This year the course changed the responsible, teacher, and examiner to Paris Carbone. The course had several important updates including enhanced visual examples, a new course project structure, the addition of automated model checking and tools (i.e., TLA+), new use-case analyses on distributed systems used at Google etc. and a new final chapter on distributed data management with a focus on distributed data processing. This year we also had to adapt to the COVID19 pandemic restrictions and compensate for the lack of physical interaction, the limited teaching environments, and focus on web-based evaluation.