A Blockchain-Based Trusted Computing Power Trading Platform

Abstract

The commercialization of distributed computing capacity is constrained by inconsistent resource descriptions, difficult-to-verify service commitments, fragmented transaction records, and limited accountability across organizations. This study develops a permissioned-ledger trading framework that combines normalized resource registration, persistent digital credentials, requirement-aware screening, contract-governed order transitions, and selective evidence anchoring. Candidate resources are first checked against mandatory conditions and are then ranked through a weighted geometric function, thereby preventing an attractive value in one attribute from concealing a serious weakness in another. Provider reliability is inferred with a regularized logistic specification that includes the interaction between certification status and observed SLA fulfillment. The framework further defines explicit rules for absent or outdated evidence, downward-biased imputation, uncertainty adjustment, and chronological model validation. Evaluation uses a reproducible dataset of 1,200 simulated providers together with an institutional computing-demand case. In the ranking experiment, the share of critically deficient resources within the highest-ranked decile declined from 35.0% under the additive baseline to 0.0% under the proposed procedure. The trust estimator increased AUC from 0.703 to 0.706 and lowered Brier loss from 0.176 to 0.165. The application case consolidated demand for 1,184 CPU cores from twelve researchers and enabled access to 1.5 PFLOPS of external computing capacity. These findings indicate that the ledger is most useful as a shared accountability and transaction-consistency mechanism, rather than as the platform on which computational workloads are executed.