부산대학교 도서관

The utilization of cloud and edge computing has become one of the most prevailing resource supply mechanisms. Thousands of enterprise users and Internet of Things (IoT) devices have performed operations through services sold by computing resource providers. So far, the auction match is the main strategy for allocating resources, in which there exists a trusted third party playing a role as a broker to deal with resource allocation requests for both providers and consumers. The main concerns in resource allocation architecture are how and how long to solve the Winner Determination Problem (WDP), which is used to lay out the match outcome. It is not easy for a single broker with limited computing power to generate an optimal solution in a short period since its corresponding time complexity is regarded as the NP-hard problem. Meanwhile, it is hard for people to trust the third party thoroughly. That is, there are three potential issues, including centralization, data security, and untrustworthiness in traditional matching architecture. In particular, the real-time matching cannot be achieved to fulfill users who have urgent needs of computing resources. To solve above issues, we have designed a trustworthy and real-time decentralized computing resource allocation platform based on blockchain and smart contract. In order to optimize the allocation results, we improve the Non-dominated Sorting Genetic Algorithm II (NSGA-II) for miners to reach the consensus mechanism. Experimental results and comparison analysis have demonstrated that potential defects could be addressed in the new method, and the real-time resource allocation can be preserved firmly despite the balance vibration of market supply and demand.