基于PoUW区块链的数据处理研究

With the advent of the era of big data, the traditional master/slave-based data processing framework can hardly meet growing computing demands. The decentralized blockchain is being considered as a completely new computing paradigm that has the potential to meet the demands. However, the blockchain is mainly designed to process transaction records, and its consensus protocol such as PoW will cause a huge waste of computing resources, as a result, the current blockchain-based framework has not been ready to become a computing paradigm. In 2017, a new consensus protocol named PoUW was proposed and overcame the aforementioned blockchain’s drawbacks at some degree. It basic idea is to let the miners mine with doing useful work (e.g., data classification). In this project, we will renovate the existing blockchain and make it become a data processing architecture. We first optimize PoUW consensus parameters so as to maximize transaction throughput and security. Then, for data center scenarios, we develop a decentralized streaming data processing architecture based on PoUW private blockchains, and fair task scheduling mechanisms for multi-resources heterogeneous environments. Finally, in order to use the idle computing resources of public networks, we will design an incentive-based shared data processing architecture, and proportional resource allocation strategies based on game theory. This project will make a crucial step toward a decentralized new computing paradigm.

随着大数据时代的来临，传统基于主/从结构的数据处理架构很难满足日益增长的计算需求。而具有去中心化结构的区块链被认为是一种极具潜力解决上述问题的全新计算范式。然而，区块链的设计初衷是用于处理交易，且所使用的共识机制如PoW会造成巨大的计算资源浪费，因此不能够被直接用于数据处理。2017年最新提出的PoUW共识机制一定程度上克服了上述缺点，其基本思想是让矿工通过做有用工作(如数据分类)来进行挖矿。本项目中，我们将改造基于PoUW的区块链使之成为一个数据处理架构。首先，我们优化PoUW共识机制参数以最大化交易吞吐量和安全性。然后，面向云数据中心，设计基于PoUW私有链的去中心化流式数据处理架构以及适用于多资源异构环境下的公平任务调度机制。最后，为利用公网的闲置计算资源，设计基于激励机制的共享数据处理架构以及基于博弈论的成比例资源分配策略。本项目的实施将为设计去中心化的全新计算范式迈出关键的一步。

常规的区块链主要用于处理交易。本项目旨在改造常规的区块链，使之成为一个一般性的数据处理架构。主要的研究任务：设计基于PoUW私有链的去中心化流式数据处理架构；设计去中心化的异构环境下的公平任务调度机制；优化PoUW共识机制参数以最大化交易吞吐量和安全性；研究资源分配策略、激励机制，在数据处理效率以及交易处理公平性方面进行折中。本项目按计划执行，实现了在论文、专利、学生培养等各方面的预期目标。本研究将为设计去中心化的全新计算范式迈出关键的一步。