Harald Ng

I am a Ph.D. student at the Data Systems Lab, advised by Paris Carbone and Seif Haridi. My research focuses on optimizing how distributed systems perform replication and transactions by leveraging practical aspects such as workloads and network settings.

Publications

Harald Ng, Seif Haridi, Paris Carbone. Omni-Paxos: Breaking the Barriers of Partial Connectivity. InEighteenth European Conference on Computer Systems (EuroSys), pages 314-330, 2023.

Harald Ng, Kun Wu, Paris Carbone. UniCache: Efficient Log Replication through Learning Workload Patterns. InInternational Conference on Extending Database Technology (EDBT), volume 26, pages 471–477, 2023.

Outreach and Development

Creator and maintainer of OmniPaxos, a distributed replication library written in Rust. OmniPaxos hides all the complexities related to consensus replication and instead provides a simple log API. This lets users build highly available services in a way that is as simple as working with a local log.

OmniPaxos is actively taught in the courses CS 651 Distributed Systems at Boston University and ID2203 Distributed Systems, Advanced Course at KTH Royal Institute of Technology.

Presentations

Invited Talks

• Leveraging Practical Aspects for Efficient Consensus, Google Sunnyvale, 2024
• Omni-Paxos: Breaking the Barriers of Partial Connectivity. Guest Lecture, CS 651 Distributed Systems, Boston University, 2023.

Conferences and Workshops

• Towards Autonomous and Pro-active State Machine Replication (Poster). In The 29th ACM Symposium on Operating Systems Principles (SOSP), 2023.
• Workload-aware Meta-Coordination for Replicated Services. In EuroSys Doctoral Workshop (EuroDW), 2023