Neonpool: Reimagining cryptocurrency transaction pools for lightweight clients and IoT devices
Hina Binte Haq, Syed Taha Ali, Asad Salman, Patrick McCorry, Siamak F. Shahandashti
TL;DR
Neonpool targets the high RAM cost of keeping unconfirmed transactions in cryptocurrency nodes by replacing full in-memory transaction storage with Bloom-filter fingerprints. It presents Neonpool-BTC and Neonpool-ETH, achieving memory reductions up to ~200x while maintaining fidelity above 99.99% and without requiring a hard fork. The approach uses ingress-based Bloom-filter filters and dstxFilters to detect duplicates/double-spends, with expiry or decaying Bloom filters to manage aging data and scaling across increasing loads. Empirical results on large Bitcoin and Ethereum datasets show favorable memory/cost trade-offs and low computation overhead on IoT-class hardware, with demonstrated adversarial resilience through per-node randomization. This work offers a practical path to enabling resource-constrained devices to participate as full nodes and improves network health and decentralization by reducing the barrier to entry for full-node operation.
Abstract
The transaction pool plays a critical role in processing and disseminating transactions in cryptocurrency networks. However, increasing transaction loads strain the resources of full node deployments. We present Neonpool, an innovative transaction pool optimization using bloom filter variants, which reduces the memory footprint of the transaction pool to a fraction. Implemented in C++ and benchmarked using a unique Bitcoin and Ethereum dataset, our solution verifies and forwards transactions with over 99.99\% accuracy and does not necessitate a hard fork. Neonpool is ideally suited for lightweight cryptocurrency clients and for resource-constrained devices such as browsers, systems-on-a-chip, mobile or IoT devices.