Proposal Distribution optimization for Endorsement Strategy in Hyperledger Fabric

High throughput and low latency are essential for the endorsement phase in the Hyperledger Fabric system (HFS). Currently, endorser nodes can be selected by clients through static configuration or service discovery to verify user-submitted proposals. However, the transparent identities of the endorser nodes in the HFS channel render them more vulnerable to attacks, and nodes with high resource consumption among the endorsers can worsen system performance. To overcome these limitations, we propose Pdo-Fabric, a proposal distribution strategy that determines the distribution weight of proposals among organizations and dynamically selects endorsers based on the resource load of nodes. This paper discusses this strategy from the following aspects: (1) the transaction latency among organizations and the selection of endorsers are investigated, together with the formalization of the problem for the endorsement phase including endorser's resource load, endorser consensus, and proposal communication latency. (2) Based on fuzzy logic, the CPU and memory resources of the endorsers are fuzzified to select endorser nodes. The number of endorser nodes is determined by a user-defined fault-tolerance probability. (3) Clients perceive the latency in processing transaction proposals and form a roulette based on the latency among multiple organizations. This roulette determines the weight of which organization will receive the distributed proposals. (4) Pdo-Fabric is implemented on top of Hyperledger Fabric platform and evaluated on metrics such as system throughput and response latency in a simulated Hyperledger Fabric environment. With a comprehensive evaluation, the proposed Pdo-Fabric system exhibits promising advancements over the existing Hyperledger Fabric endorsement strategy.