João Oliveira

My name is João Oliveira, I’m a PhD student at Instituto Superior Técnico at Universidade de Lisboa, doing research at INESC-ID in the Distributed Systems Group. My advisor is Miguel Matos.

My research topic is titled Performance and Correctness of Emerging Persistent Memory Systems. Currently I’m interested in the possibilities brought forth by the Compute eXpress Link (CXL) for the development of correct and scalable distributed systems.

news

Apr 03, 2025	I attended Eurosys 2025 where I presented our work. HawkSet received the EuroSys Gilles Muller Best Artifact Award.
Mar 01, 2025	Our paper HawkSet: Automatic, Application-Agnostic, and Efficient Concurrent PM Bug Detection was accepted at Eurosys 2025.
May 01, 2024	Participated in the combined Artifact Evaluation Committee for ATC and OSDI.
Nov 25, 2023	I submitted and sucessfully defended my Msc Dissertation at Instituto Superior Técnico, and received a mark of 20/20.
Sep 07, 2023	I attended INForum 2023 in Porto, where I presented my current work-in-progress in the form of a live presentation and a poster.

selected publications

EuroSys2025

HawkSet: Automatic, Application-Agnostic, and Efficient Concurrent PM Bug Detection

João Oliveira, João Gonçalves, and Miguel Matos

In Proceedings of the Twentieth European Conference on Computer Systems, 2025

Abs DOI

Persistent Memory (PM) enables the development of fast, persistent applications without employing costly HDD/SSD-based I/O operations. Since caches are volatile and CPUs may reorder and stall memory accesses for performance, developers must use low-level instructions to ensure a consistent state in case of a crash. Failure to do so can result in data corruption, data loss, or undefined behavior. In concurrent executions, this exposes a new class of bugs. HawkSet is an automatic, application-agnostic, and efficient tool to detect concurrent PM bugs. HawkSet uses lockset analysis, and automatic binary instrumentation to find all the bugs detected by the state-of-the-art tools and 7 previously unknown bugs. This is achieved without requiring application-specific knowledge or models, nor specialized debugging artifacts or guided executions. Compared to the state-of-the-art, HawkSet offers up to a 159x speedup, and consistently detects harder-to-reach bugs, where a rare interleaving is required.