Keynote Speakers:

  • Yair Amir, Johns Hopkins University, USA

    • Slides from the presentation.

  • Biography:

    Yair Amir is a Professor Emeritus of Computer Science, and director of the Distributed Systems and Networks lab at Johns Hopkins University.

    Dr. Amir served as department chair of Computer Science, Johns Hopkins (2015-2018), as general co-chair (2022) and program co-chair (2015) of the IEEE/IFIP Dependable Systems and Networks (DSN) conference, and as a Vice Chair of the IFIP 10.4 Working Group on Dependable Computing (2016-2018).

    He is a creator of several open-source technologies, including the Spire intrusion-tolerant SCADA for the power grid, the Spines overlay network platform, and the Spread toolkit. Some of these technologies are deployed in mission critical and are used for research and teaching around the world.

    Dr. Amir is a co-founder of Spread Concepts LLC and a co-founder of LTN Global Communications. LTN offers global video transport and processing services that are used by major broadcasters including Disney, YouTube TV, Fox, CNN, ABC, Bloomberg, CBS, ESPN, NBC, PBS, and Turner.

    He holds B.Sc. and M.Sc. from the Technion, Israel Institute of Technology, and a Ph.D from the Hebrew University of Jerusalem, Israel.

  • Title: Persistent Control Plane for a Global Cloud - from an Abstract Algorithm to a Deployed Service
  • Abstract:

    The LTN cloud is a specialized global cloud providing real-time, broadcast-quality video transport and processing services supporting diverse workflows for the media industry (LTN Global Communications). The system is governed by a persistent control plane that relies on database replication with strict consistency semantics, while also adhering to stringent latency requirements. Notably, the end-to-end latency, from the moment an update is initiated anywhere in the world to the point at which it is reflected globally, is approximately a quarter of a second.

    The talk describes the path from an abstract algorithm to a deployed service. This includes (1) the overlay network and group communication architecture that effectively provides a near-perfect communication service over the asynchronous and error-prone Internet, (2) the algorithmic refinements that were necessary to support the latency requirement in the face of server recoveries and network partition repairs, and (3) a thoughtful setup of data centers hosting replicas, striking a balance between strong resilience and low-latency total ordering.

    Since its deployment in 2019, the persistent control plane service has experienced a three orders of magnitude increase in the volume of updates. This growth is attributed to the broader adoption of the service as an essential building block for various LTN cloud applications, surpassing its original intended use case.

Invited Papers:

    Regular:

    • Common Public Knowledge for Enhancing Machine Learning Data Sets, Arnon Ilani (Ben-Gurion University of the Negev, Beersheva, Israel) and Shlomi Dolev (Ben-Gurion University of the Negev, Beersheva, Israel). Slides from the presentation.
    • Lessons from HotStuff, Dahlia Malkhi (Chainlink Labs) and Maofan Yin (Chainlink Labs). Slides from the presentation.
    • Disaggregating Applications Using Uniservices, Xinwen Wang (Cornell University) and Robbert van Renesse (Cornell University). Slides from the presentation.
    • Initial Steps Toward a Compiler for Distributed Programs, Joseph Hellerstein (University of California, Berkeley), Shadaj Laddad (University of California, Berkeley), Mae Milano (University of California, Berkeley), Conor Power (University of California, Berkeley), and Mingwei Samuel (Sutter Hill Ventures). Slides from the presentation.
    • Oblivious Transfer Protocol without Physical Transfer of Hardware Root-of-Trust, Boyapally Harishma (Indian Institute of Technology Kharagpur), Chandan Kumar Chaudhary (Indian Institute of Technology Kharagpur), and Debdeep Mukhopadhyay (Indian Institute of Technology Kharagpur). Slides from the presentation.

    Short research statements:

    • ServerFilling: A better approach to packing multiserver jobs, Isaac Grosof (Carnegie Mellon University) and Mor Harchol-Balter (Carnegie Mellon University).
    • Towards Efficient Microservice Communication, Divyanshu Saxena (UT Austin), William Zhang (UT Austin), Madhav Tummala (UT Austin), Saksham Goel (UT Austin), and Aditya Akella (UT Austin and Google).
    • Planetary Scale Byzantine Consensus, Gauthier Voron (Ecole Polytechnique Fédérale de Lausanne, EPFL) and Vincent Gramoli (University of Sydney). Slides from the presentation.
    • Enhancing Privacy in Federated Learning via Early Exit, Yashuo Wu (University of California, Irvine), Carla Fabiana Chiasserini (Politecnico di Torino), Francesco Malandrino (CNR-IEIIT), and Marco Levorato (University of California, Irvine).

Peer-reviewed Papers:

    Regular:

    • An Extensible Framework for Implementing and Validating Byzantine Fault-Tolerant Protocols, Hanish Gogada (University of Stavanger), Hein Meling (University of Stavanger), Leander Jehl (University of Stavanger), and John Ingve Olsen (University of Stavanger).
    • Specification and Runtime Checking of Derecho, A Protocol for Fast Replication for Cloud Services, Kumar Shivam (Stony Brook University), Vishnu Paladugu (Stony Brook University), and Yanhong A. Liu (Stony Brook University). Slides from the presentation.

    Short research statements:

    • Performance of EdDSA and BLS Signatures in Committee-Based Consensus, Zhuolun Li (University of Leeds), Alberto Sonnino (Mysten Labs, UCL), and Philipp Jovanovic (UCL).
    • ParSwarm: A C++ Framework for Evaluating Distributed Algorithms for Robot Swarms, Zhi Wei Gan (Massachusetts Institute of Technology), Grace Cai (Massachusetts Institute of Technology), Noble Harasha (Massachusetts Institute of Technology), Nancy Lynch (Massachusetts Institute of Technology), and Julian Shun (Massachusetts Institute of Technology). Slides from the presentation.