Abstract
The growing ubiquity of Blockchain has led to the experimental transplantation of classical web applications into the Blockchain space. Notoriously, running computationally-expensive processes on-chain is less than ideal. Therefore, it is crucial to find ways to utilize the security and decentralization provided by Blockchain technology while reducing on-chain computational load. To that end, recent work has shown the benefit of Multi-stage Transaction architectures. The benefit of such systems is that they provide requested data quickly to the client while still ensuring that accurate results are eventually returned. In this paper, we describe a prototype of a multi-stage transaction system, which we call Friend in Mobile Edge (FiME), that intelligently coordinates client requests between a Blockchain and a central cloud node. We select the use case of stock price prediction and evaluate based on response time to the client as well as computational load on the various components. Our results involve an analysis of the performance of such a system in delivering timely results to the end user, as well as a discussion of the drawbacks of handling large data requests on-chain. We also include recommendations for how future systems can build on these findings.