Xu et al. - 2019 - Making Big Data Open in Edges A

文件名称: Xu et al. - 2019 - Making Big Data Open in Edges A Resource-Efficien.pdf

所属分类: 比特币

开发工具:

文件大小: 1mb

下载次数: 0

上传时间: 2019-10-05

提供者: liang******

下载 (1mb)

不能下载？报告错误

详细说明：论文 - Making Big Data Open in Edges: A Resource-Efficient Blockchain-Based Approach872 lEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS. VOL 30. NO. 4. APRIL 2019 mobile edge computing collaboration in 5G ecosystem. The and the throughput can achieve up to tens of thousands of authors illustrated a context-aware and dynamic collabora- transactions per second. Milutinovic et al. [27] designed a tion infrastructure in the edge of Radio Access Network time-and energy- efficient blockchain consensus algorithm (RAN consisting of mobile edge devices, edge services and based on extra trusted execution environments. The core base stations, where the heterogeneous resources are merged function of this algorithm is constructed on the random It edges. Zhang et al. [14] developed a novel computing par- number generation of the underpinned trusted execution adigm for big data sharing called Firework in collaborative environment. Turesson et al. [28] proposed to use deep edges, where virtual shared data views are built and data is learning model training rather than calculate hash in Proof transmitted to users through predefined interfaces. This of-Work consensus Luu et al. [29] proposed a secure shard framework guarantees users' privacy as well as solving the ing protocol for blockchain mining, which is also used in response latency issue by pushing data to the network edges. Ethereum for improving the throughput and scalability Wu et al. [15] proposed a two-step detection mechanism in Zheng et al. [30] gave an overview of classic blockchain mobilc cdge collaboration, whcrc uscrs preferences arc con- technologics from the aspects of thc architccture and the cerned for constructing virtual communities and collabora- consensus Based on this work, Bach et al. 131 performed a tive clusters. Moreover, a video coding sharing mechanism further comparative analysis of several classic and modern based on users identities is developed for flexible video dis- blockchain consensus algorithms from complexity, scalabil- tribution and decreasing energy consumption at the edge of ity, security, and rewarding method mobile networks. To adapt blockchain to the edge comput Different from these previous works, we first develop a ing, Xiong et al. [16], [17l proposed to offload the mining task green blockchain framework to enable trust for big data to the edge computing service providers, who make profit sharing in collaborative edges. Then, we put forward green by providing computational resource. In the proposed sce- PoC consensus mechanism in our framework to reduce nario, Stackelberg game is used to optimize the price of the computational resources in edges, where edge devices give resource Stanciu [11] proposcd to usc blockchain as a dis- thcir proof of contributing collaboration to compete for tributed control system confirming to the IEC standard. block generation, rather than wasting computational resour Samaniego et al. [181, [19 leveraged blockchain as the carrier ces to solve mathematic puzzle. Furthermore, we propose a of the virtual resource, which is a kind of micro-services, to futile transaction theory and establish transaction offloading reduce the computation moving cost on edge hosts module based on ftf algorithm for reduction of storage resources occupied by blockchain. Finally, we design 2.2 Blockchain Technology Express Transaction and Hollow Block to reduce the usage Blockchain technology has aroused great interests from both of the network resource in blockchain academic and industrial fields, including finance, e-health, distributed system, etc. Christidis et al. [20] presented a 3 FRAMEWORK DESIGN comprehensive survey on blockchain and claimed that the Edges consist of edge infrastructures, base stations, edge serv blockchain can be employed to construct a resilient distrib- ers, and loT edge devices, etc. Every edge links to the network uted system in which participants could interact with each served by different Internet Service Providers(ISPs). In our other without a trusted third party. They demonstrated that proposal, we deploy blockchain on these edges, where ever the combination of blockchain and loT can make significant block contains multiple transaction logs of big data flows improvements. Azaria et al. 121] designed a blockchain among edge applications. For a more clear description of based system called MedRec for electronic medical record proposal, we demonstrate a green blockchain framework in management. In the MedRec, medical stakeholders such as collaborative edges in this section. Our proposed framework medical scientist and public health authorities are involved is divided into four layers, as shown in Fig. 1 as miners. Weber et al. [22] adapted blockchain for business. AI PI layer offers interfaces for edge applications, which The trust of blockchain underpins the international business abstract the functions of cache and blockchain layer to pro- process. The authors performed three case studies to illus- vide various calls for implementing edge collaboration. Sp trate the feasibility of their proposed solution cifically, api layer contains following operations here were several works focusing on the inner mecha- nism of blockchain technologies Saito et al. [23] proposed Read, write, and execute operations abstract transac that blockchain can be regarded as a probabilistic state tion construction in the blockchain layer machine, where the amount of participants is uncertain and Policy configuration is designed to set the operation participants cannot make commitment on the decisions permission to local data for other edge devices Eyal et al. [24] developed a novel Bitcoin-NG (Next Genera Query operation can query operation record of other tion)protocol to improve the scalability, which belongs to a edge devices on local data, where the latest opera- kind of Byzantine fault tolerant protocol. In addition, a set tions of local data are stored in the local operation of metrics standard is introduced to quantify the security modle in cache layer. and efficiency of the blockchain protocols. Lewenberg et al The cache layer is designed to accelerate the responses to [25] designed a Directed Acyclic Graph (DAg)structure for the calls, and it contains local operations, invalidated blocks blockchain to enhance the throughout, where a block could and useful blocks reference many predecessors. Miller et al. [26] proposed a The blockchain layer implements the content of block practical asynchronous Honey Badger BFT protocol. This chain in edges, including several modules as follows protocol can ensure normal operation without any time First, transaction and block construction module assumptions on a wide area network more than 100 nodes transforms the requests from the upper layer into XU ETAL. MAKING BIG DATA OPEN IN EDGES: A RESOURCE-EFFICIENT BLOCKCHAIN-BASED APPROACH entire edge network for validation quite less computational resources than that of Pol requires transactions or blocks, which will be broadcast to the mining. It is a promising replacer of Pow, since it Second, transaction validation module contains vali In addition, Practical Byzantine Fault Tolerant(PBFT) dation rules, where the operation permission to local and its variants are widely used in consortium chains, data is often set for other edge devices via modifying which tolerate up to a third of participants that occur any validation rules. Besides, transaction and block vali- form of failure(Byzantine fault), given the number of partic- dation modules guarantee rules, which are founda- ipants in advance and fixed [381 tions of the green PoC consensus, as we will Within the context of collaborative edges, as mentioned illustrate in Section 4 above, every edge device is a participant of the network, Finally, transaction offloading module locates the and may require to perform blockchain operations More blocks with useful transactions, and then the useful over, the number of edge devices, which should adapt to blocks are updated to the cache layer. This module is the demand of users, is not fixed. As we mentioned in dcsigned to rcducc storage resources occupied by Scction 1, thc blockchain bascd cdge collaboration urges to blockchain pursue a green solution because of the limited computa- Storage layer in the bottom provides persistent storage tional and storage resources. Hence, inspired by pos and ervice for the upper layers PoW, we will illustrate Poc in details in next subsection 4 GREEN POOF-OF-COLLABORATION CONSENSUS 4.2 Proof-of-Collaboration Mechanism MECHANISM Edge devices give the proof of their contribution to collabo ration rather than solvc meaningless mathcmatical puzzle Blockchain is a distributed data structure and every partici pant keeps a copy of the entire blockchain [32 ,[33],[] to obtain the privileges of collaboration the key concept of The first class component in blockchain is named transac- Proof-of-Collaboration is that participants contribute to the big data sharing so that they can also benefit from other par tion, which is a record of some asset transferring. These ticipants collaboration. More specifically, the green poc transactions generated by different devices are validated via a whole blockchain network, and are packaged into a consensus mechanism is designed as followS block by a miner. Then miners keep consistency of blocks validation via performing consensus mechanism. Finally 4.2.1 Collaboration Credit valid block is added to the blockchain In our design, the edge collaboration is underpinned by a new assct called Collaboration Credit(CC), which is slightly similar 4.1 Different Chain Types and Consensus to btC in Bitcoin 33 and ETH (GaS)in Ethereum 39 This The blockchain has two types: a public chain and a conson means that the data flow from edge applications recorded by tium(private) chain. If anyone can participate in a bloc transactions, i.e., collaborations, must be paid using CC in the chain network the blockchain is naturally public or proposed framework. The CC used for this payment is dynam consortium 3 65]. If every participant can take part in block- ically dctcrmincd by collaboration fcc F as chain operations, for example, competing to mine blocks or proposing transactions, the blockchain is public [20]. Since the public chain is open and competitive, the participants in y CC/kB ×m ublic chain nctwork do not trust cach other 36 On the contrary, the participants of a consortium chain network are where y/ is a pre-defined throughput threshold, y/'represents privileged and white-listed the average throughput of the entire network during recent The differences between the two chain types result in dif- 100 blocks, and n denotes the number of edge devices in the ferent kinds of potential applied consensus protocols. Giv- network. The average network throughput y can be calcu ing any participant an opportunity to minc blocks, Proof-of- lated by dividing the total size of transactions in recent 100 Work (pow) makes a great success in Bit soin H hich is the blocks by the time consumption of generating these 100 iggest public chain in the world 133, 134]. Pow requires blocks. In practice, v equals to the maximum value of devices' articipants that compete for mining blocks to give the network capacity. according to the definition of F, the frame proof of their work. This proof is a kind of mathematical work will decrease to encourage collaboration when the puzzle that is casy to bc validated but cxtremcly hard to be recent throughput is lower than the pre-defined threshold,or solved, i. e, solving these kinds of puzzles consumes fabu- increase F to reduce network overload when the throughput lous amount of computational resources. In most cases, the s higher than defined. Moreover, the larger the amount of puzzle has the following form edge devices is, the lower F will be in the framework Find n the framework, CC can be gained by two approaches (1) First, the block proposer can be rewarded a certain number S.t. SHA256(SHA256(.m))< target, of Cc by adding a new block to the blockchain successfully Second, the block proposer earns CC from the transactions here. is a string concatenate operator, and h represents carried by the block. The collaboration fee F is used to eval the content of the newest block. The smaller the target is the uate the contribution, i. e to prevent selfish applications more difficult the mining is. Later, the concept of Proof-of- requesting shared data without sharing their own data. If Stake(pos)[37] has been proposed, and its main idea is that an edge application leverages data flow from other applica stakeholders should show their stake of assets to compete tions, it must contribute to the edge collaboration lEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS. VOL 30. NO. 4. APRIL 2019 4.2.2 Proof-of-Collaboration Legen In the framework, the way to propose a block is rclated to O Useful block the persistence p which is defined as the time since the last CC changes Our proposal has the following three core rules, □ Useful transaction Inout I Futile transaction underpinned by CC and P, to guarantee itself a green blockchain Rule 1 dynamic difficulty). The mining in the proposed PoC is different from Eq. (1). Mining in PoC is influenced by dynamic difficulty, which is different from various participants It has the form as follows st.SHA256(SHA256(h.m)月 3210 0 40 20 Block Heigh ght (thousand) Blockchain Height(thousand) Fig 8 Accumulative energy. Fig. 10. Comparison of storage cost 1e14 E-TX+HB 1600 E-TX 1.2 PoC PoW 61400 Traditional 1.0 0.8 200 s0.6 1000 0.4 800 0.2 600 0.0 01000020000300004000050000 0 20 30 Block Size(transactions/Block) Block Height( thousand) Fig. 11. Comparison of throughput Fig 9. Accumulative hashes 300 Traditional e storage cost grows linearly without transaction offload- 艺250 HB ing. Under the control of this offloading module, the storage 200 cost of our green blockchain grows slowly and is stabilized t a low level. This is because the proposed module can rec 第150 ognize the transactions that cannot be further referenced 100 and upload them to the cloud for reducing storage We compare the network throughput of traditional 50 blockchain and blockchain with F-TX and hollow block indicated by IIB)in Fig. 11. The throughput of blockchain with E-TX is 17 percent higher than traditional blockchain Hollow block further enhances the throughput to 23 percent Number of transactions higher than traditional blockchain. The enhancement is Fig 12. Comparison of network usage mainly contributed by the asynchronous validation of f-tX and the redundancy reduction provided by hollow block To show the capability of redundancy reduction we 7.3 Discussion demonstrate the network usage of recording transactions Security limitation. The proposed PoC mechanism reduces into blockchain with hollow block. Fig. 12 shows that the the waste of computational resource, enabling blockchain hollow blocks reduce about 27 percent network usage of a based edge collaboration. Previous work indicates that the single participant by removing the redundancy between security of the Pow protocol is based on the"wasted"com participant's queue and new blocks putation [55]. However, this assert is based on the threat During the experiment, the average latency of submitting model that the adversary aims to control the whole system a transaction is 38 ms, which is lower than the requirement This model considers the blockchain system as an entity, of latency(100 ms [52) in 5G edge and IoT network [531, which actually is an extreme case. In more general models, [54]. For each data sharing operation, the proposed method the adversary is able to start a double-spending attack if he will bring 290 Bytes of communication overhead on aver- owns more computational resource than any of the partici- age. In summary, the results of experiments show that our pants. In this situation, the security performance is deter proposed green blockchain is able to reduce enormous mined by the participant owning the most computational computational, storage, and network resources, which helps resource. Therefore, the security level of Pow and Poc in to solve the critical challenges we describe in Section 1 neral cases can be the same

(系统自动生成,下载前可以参看下载内容)