Two Papers Win the 2025 Jean Claude Laprie Award

For 2025, the Award Committee has unanimously decided to select the following two papers:

Song Liu, Karthik Pattabiraman, Thomas Moscibroda, and Benjamin G. Zorn. 2011. Flikker: saving DRAM refresh-power through critical data partitioning. In Proceedings of the sixteenth international conference on Architectural support for programming languages and operating systems (ASPLOS XVI). Association for Computing Machinery, New York, NY, USA, 213–224. https://doi.org/10.1145/1950365.1950391

Patrick Eugster, Rachid Guerraoui, S. B. Handurukande, Petr Kouznetsov, and Anne-Marie Kermarrec. 2001. Lightweight Probabilistic Broadcast. In Proceedings of the 2001 International Conference on Dependable Systems and Networks (DSN 2001). IEEE Computer Society, USA, 443–452, https://doi.org/10.1109/DSN.2001.941428.

The Award Citation:

“Flikker: saving DRAM refresh-power through critical data partitioning” by Song Liu, Karthik Pattabiraman, Thomas Moscibroda, and Benjamin G. Zorn, was one of the first papers in the field of approximate computing, and attempted to bridge the worlds of dependability and energy efficiency (i.e., sustainability). The main idea was to trade off hardware reliability to save power in DRAM memories, and to allow programmers to control the tradeoff based on the semantics of their applications. While other papers had also explored the power-reliability tradeoff, Flikker was the first to do so at the programming language level. This was a fundamental innovation, as it allowed fine-grained control over the application’s semantics and quantified the risk that the developer was prepared to tolerate for achieving energy savings. Many other papers have built on the ideas proposed in this paper, as evidenced by the fact that it has been cited more than 650 times since it’s publication in 2011. The paper revitalized research into DRAM power savings, especially for data centers (though the paper’s focus was on mobile computing platforms, the ideas translate to data centers and high-performance computing as well). This is important as data centers are one of the leading causes of global warming today, and are projected to grow substantially in the near future due to the need for compute power in training and inference of AI workloads. The ideas proposed in the paper have been widely adopted in industry. In particular, Microsoft has used similar techniques for their search cluster (Bing), and other companies have also invested in programmatically controlling DRAM power consumption at the application level. In the 15 years since the paper’s publication, DRAM manufacturers have incorporated even more power-savings modes in their DRAM modules (at the time of the paper’s publication, only two modes were available), which makes the Flikker idea even more applicable to modern DRAMs. For these reasons the award Committee decided to select this paper as one of the 2025 JCL award winners”.

“Lightweight probabilistic broadcast” by Patrick Th. Eugster, Rachid Guerraoui, S. B. Handurukande, Petr Kouznetsov, and Anne-Marie Kermarrec, presented the first decentralised, reliable information dissemination protocol. The protocol puts to work the intuitive idea of gossip, or epidemic information dissemination, in a completely decentralised setting without any single point of failure. The original idea, consisting of transmitting news items in a peer-to-peer manner by periodically having each peer pick the one to which to transmit the item randomly among its neighbours, had been considered quite appealing since the work of Demers et al. on replicated databases in the 90s. However, until this paper, all approaches assumed a trusted global knowledge about the participants in a static system. The reliability of the decentralised algorithm presented in the nominated paper is formally analysed using a stochastic approach. The paper shows that the view size has very little impact on the latency of message dissemination. Interestingly, the paper also shows that the probability of partition creation in the system decreases as the system grows in size. Empirical evaluations support the formal analysis. The paper has been cited more than 800 times, it has impacted the design of many systems (e.g. Scribe, Splistream, Bullet, Cyclon, Tman, Scamp, BAR Gossip, Hermes, Spidercast, Tera and many others), on systems spanning a large class of applications: vehicular control (including cars and drones), live streaming, data storage (in a practical a large-scale storage system used in CERN), gaming, smart grids, NATS, Manets, blockchain systems, editing systems, digital libraries, healthcare and software-defined networks. For these reasons, the award Committee decided to also select this paper as one of the 2025 JCL award winners”.

This entry was posted in JCL award. Bookmark the permalink.