eAID: Elastic Asynchronous Information Dispersal with Post-Dissemination Pruning
🗞️ “eAID: Elastic Asynchronous Information Dispersal with Post-Dissemination Pruning” will appear at the coming DISC’26 conference.
👿 Distributed fault tolerant algorithms face a double-adversary: the network may delay responses from F participants indefinitely; and up to F of those respondents may later become faulty. This conservative design often leads to suboptimal communication and storage costs.
📢 eAID performs a dynamic assessment of how many responses are gathered, often more than N-F, and only up to F of those respondents may later become faulty.
The results are substantial improvements in latency, communication overhead, and storage costs. For a preprint, see [https://arxiv.org/abs/2603.24761].
Authors: Rithwik Kerur, Dahlia Malkhi, Divyakant Agrawal, Mike Reiter, Amit Wieder.