Pub Date : 2021-12-01DOI: 10.1109/ICPADS53394.2021.00066
Yangfei Lin, Jie Li, Shigetomo Kimura, Yongbing Zhang, Yusheng Ji, Yang Yang
Cloud storage services offer flexible, convenient solutions for business and personal users to store data. Traditionally, Third Party Auditors (TPAs) are introduced to ensure data integrity for public auditing. However, TPAs may also be untrusted for forging the auditing results or colluding with cloud storage servers to deceive users. In this paper, we propose a novel Blockchain-based Public Auditing Outsourcing system without TPAs (BPAO), in which the computationally expensive operations in public auditing are outsourced through blockchain to the cloud servers without risking users' privacy. Our security analysis indicates that BPAO achieves soundness and robustness. The experimental results show that BPAO is computationally efficient for cloud storage user.
{"title":"Blockchain based Public Auditing Outsourcing for Cloud Storage","authors":"Yangfei Lin, Jie Li, Shigetomo Kimura, Yongbing Zhang, Yusheng Ji, Yang Yang","doi":"10.1109/ICPADS53394.2021.00066","DOIUrl":"https://doi.org/10.1109/ICPADS53394.2021.00066","url":null,"abstract":"Cloud storage services offer flexible, convenient solutions for business and personal users to store data. Traditionally, Third Party Auditors (TPAs) are introduced to ensure data integrity for public auditing. However, TPAs may also be untrusted for forging the auditing results or colluding with cloud storage servers to deceive users. In this paper, we propose a novel Blockchain-based Public Auditing Outsourcing system without TPAs (BPAO), in which the computationally expensive operations in public auditing are outsourced through blockchain to the cloud servers without risking users' privacy. Our security analysis indicates that BPAO achieves soundness and robustness. The experimental results show that BPAO is computationally efficient for cloud storage user.","PeriodicalId":309508,"journal":{"name":"2021 IEEE 27th International Conference on Parallel and Distributed Systems (ICPADS)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125667821","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-01DOI: 10.1109/ICPADS53394.2021.00095
Qian Zhang, Sheng Cao, Xiaosong Zhang
Nowadays, online medical services have been greatly developing. Cryptocurrencies like Bitcoin and Ethereum are very suitable for online medical electronic payment scenarios that require identity privacy protection because of their good anonymity and financial payment attributes. However, cryptocurrencies varies widely, the need of cryptocurrencies exchange is urgent for patients to pay different doctors and platforms with diverse cryptocurrencies. Exchanging cryptocurrencies through centralized exchanges has problems such as high fees and cumbersome operations. The decentralized exchanges mainly focus on cross-blockchain connectivity but high intermediate fees charged by connectors are ignored. In order to minimize the exchange fees, we propose a cross-blockchain connector selection scheme utilizing the reverse Vickrey auction along with Interledger. Our scheme abstracts the connector nodes selection into a service provider bidding process, throught which we can find the very node with the lowest bid, namely, the least exchange fees, as the ideal cross-blockchain service provider. Our scheme implements cross-blockchain payment of different cryptocurrencies conveniently, quickly and cheaply, which can provide patients with better identity protection of personal privacy information. Security analysis and performance evaluations show that our scheme can effectively promote the applications of cryptocurrencies in the field of medical care.
{"title":"Enabling Auction-based Cross-Blockchain Protocol for Online Anonymous Payment","authors":"Qian Zhang, Sheng Cao, Xiaosong Zhang","doi":"10.1109/ICPADS53394.2021.00095","DOIUrl":"https://doi.org/10.1109/ICPADS53394.2021.00095","url":null,"abstract":"Nowadays, online medical services have been greatly developing. Cryptocurrencies like Bitcoin and Ethereum are very suitable for online medical electronic payment scenarios that require identity privacy protection because of their good anonymity and financial payment attributes. However, cryptocurrencies varies widely, the need of cryptocurrencies exchange is urgent for patients to pay different doctors and platforms with diverse cryptocurrencies. Exchanging cryptocurrencies through centralized exchanges has problems such as high fees and cumbersome operations. The decentralized exchanges mainly focus on cross-blockchain connectivity but high intermediate fees charged by connectors are ignored. In order to minimize the exchange fees, we propose a cross-blockchain connector selection scheme utilizing the reverse Vickrey auction along with Interledger. Our scheme abstracts the connector nodes selection into a service provider bidding process, throught which we can find the very node with the lowest bid, namely, the least exchange fees, as the ideal cross-blockchain service provider. Our scheme implements cross-blockchain payment of different cryptocurrencies conveniently, quickly and cheaply, which can provide patients with better identity protection of personal privacy information. Security analysis and performance evaluations show that our scheme can effectively promote the applications of cryptocurrencies in the field of medical care.","PeriodicalId":309508,"journal":{"name":"2021 IEEE 27th International Conference on Parallel and Distributed Systems (ICPADS)","volume":"39 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130709349","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-01DOI: 10.1109/ICPADS53394.2021.00086
Y. Zhang, D. Goswami
Research on Graph Convolutional Networks (GCNs) has increasingly gained popularity in recent years due to the powerful representational capacity of graphs. A common assumption in traditional synchronous parallel training of GCNs using multiple GPUs is that load is perfectly balanced. However, this assumption may not hold in a real-world scenario where there can be imbalances in workloads among GPUs for various reasons. In a synchronous parallel implementation, a straggler in the system can limit the overall speed up of parallel training. To address these performance issues, this research investigates approaches for asynchronous decentralized parallel training of GCNs on a GPU cluster. The techniques investigated are based on graph clustering and the Gossip protocol. The research specifically adapts the approach of Cluster GCN, which uses graph partitioning for SGD based training, and combines with a gossip algorithm specifically designed for a GPU cluster to periodically exchange gradients among randomly chosen partners (GPUs). In addition, it incorporates a work pool mechanism for load balancing among GPUs. The gossip algorithm is proven to be deadlock free. The implementation is performed on a deep learning cluster with 8 Tesla V100 GPUs per compute node, and PyTorch and DGL as the software platforms. Experiments are conducted on different benchmark datasets. The results demonstrate superior performance with similar accuracy scores, as compared to traditional synchronous training which uses “all reduce” to synchronously accumulate parallel training results.
{"title":"Efficient Asynchronous GCN Training on a GPU Cluster","authors":"Y. Zhang, D. Goswami","doi":"10.1109/ICPADS53394.2021.00086","DOIUrl":"https://doi.org/10.1109/ICPADS53394.2021.00086","url":null,"abstract":"Research on Graph Convolutional Networks (GCNs) has increasingly gained popularity in recent years due to the powerful representational capacity of graphs. A common assumption in traditional synchronous parallel training of GCNs using multiple GPUs is that load is perfectly balanced. However, this assumption may not hold in a real-world scenario where there can be imbalances in workloads among GPUs for various reasons. In a synchronous parallel implementation, a straggler in the system can limit the overall speed up of parallel training. To address these performance issues, this research investigates approaches for asynchronous decentralized parallel training of GCNs on a GPU cluster. The techniques investigated are based on graph clustering and the Gossip protocol. The research specifically adapts the approach of Cluster GCN, which uses graph partitioning for SGD based training, and combines with a gossip algorithm specifically designed for a GPU cluster to periodically exchange gradients among randomly chosen partners (GPUs). In addition, it incorporates a work pool mechanism for load balancing among GPUs. The gossip algorithm is proven to be deadlock free. The implementation is performed on a deep learning cluster with 8 Tesla V100 GPUs per compute node, and PyTorch and DGL as the software platforms. Experiments are conducted on different benchmark datasets. The results demonstrate superior performance with similar accuracy scores, as compared to traditional synchronous training which uses “all reduce” to synchronously accumulate parallel training results.","PeriodicalId":309508,"journal":{"name":"2021 IEEE 27th International Conference on Parallel and Distributed Systems (ICPADS)","volume":"83 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114255170","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-01DOI: 10.1109/ICPADS53394.2021.00069
Lirui Zhao, Jipeng Zhang, Junhao Huang, Zhe Liu, G. Hancke
Kyber, an IND-CCA-secure key encapsulation mechanism (KEM) based on the MLWE problem, has been shortlisted for the third round evaluation of the NIST Post-Quantum Cryptography Standardization. In this paper, we explored the optimizations of Kyber in high-performance processors from the ARM Cortex-A series, which are widely used in mainstream mobile phones. To improve the performance of Kyber, we utilized the powerful SIMD instruction set NEON in an ARMv8-A to parallelize the core modules of Kyber, i.e., modular reduction and NTT. Specifically, we specially designed the optimized implementation based on the characteristic of the NEON instruction set for the Barrett and Montgomery reduction algorithms. To make full use of the computing power of NEON instructions, we proposed a novel strategy for computing the 16-bit Barrett reduction without handling the 32-bit intermediate result. Our Barrett and Montgomery reduction showed 8.52 and 8.89 times faster than the reference implementation. As for NTT/INTT, we adopted the 2+5 layer merging strategy on an ARMv8-A to implement NTT/INTT after carefully analyzing the register occupancy of various layer merging techniques. Thanks to the selected layer merging strategy, our NTT and INTT achieved 11.89 and 13.45 times speedups compared with the reference implementation. Our optimized software achieved 1.77×, 1.85×, and 2.16× speedups for key generation, encapsulation, and decapsulation compared with Kyber's reference implementation.
{"title":"Efficient Implementation of Kyber on Mobile Devices","authors":"Lirui Zhao, Jipeng Zhang, Junhao Huang, Zhe Liu, G. Hancke","doi":"10.1109/ICPADS53394.2021.00069","DOIUrl":"https://doi.org/10.1109/ICPADS53394.2021.00069","url":null,"abstract":"Kyber, an IND-CCA-secure key encapsulation mechanism (KEM) based on the MLWE problem, has been shortlisted for the third round evaluation of the NIST Post-Quantum Cryptography Standardization. In this paper, we explored the optimizations of Kyber in high-performance processors from the ARM Cortex-A series, which are widely used in mainstream mobile phones. To improve the performance of Kyber, we utilized the powerful SIMD instruction set NEON in an ARMv8-A to parallelize the core modules of Kyber, i.e., modular reduction and NTT. Specifically, we specially designed the optimized implementation based on the characteristic of the NEON instruction set for the Barrett and Montgomery reduction algorithms. To make full use of the computing power of NEON instructions, we proposed a novel strategy for computing the 16-bit Barrett reduction without handling the 32-bit intermediate result. Our Barrett and Montgomery reduction showed 8.52 and 8.89 times faster than the reference implementation. As for NTT/INTT, we adopted the 2+5 layer merging strategy on an ARMv8-A to implement NTT/INTT after carefully analyzing the register occupancy of various layer merging techniques. Thanks to the selected layer merging strategy, our NTT and INTT achieved 11.89 and 13.45 times speedups compared with the reference implementation. Our optimized software achieved 1.77×, 1.85×, and 2.16× speedups for key generation, encapsulation, and decapsulation compared with Kyber's reference implementation.","PeriodicalId":309508,"journal":{"name":"2021 IEEE 27th International Conference on Parallel and Distributed Systems (ICPADS)","volume":"50 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116507542","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-01DOI: 10.1109/ICPADS53394.2021.00050
Yuanhe Shu, Jingwei Wang, L. Kong, Jiadi Yu, Guisong Yang, Yueping Cai, Zhen Wang, M. K. Khan
The booming of mobile technologies and Internet of Things (IoTs) have facilitated the explosion of wireless devices and brought convenience to people's daily lives. Coming with the explosive growth of wireless devices, incompatibility of heterogeneous wireless technologies hindered the growing demands for everything connected. And spectrum sharing among heterogeneous wireless technologies has led to severe Cross-Technology Interference (CTI), which is a vital obstacle for network reliability and spectrum utilization. Researches in recent years have shown that Cross-Technology Communication (CTC) turns out to be a promising solution with broad perspective for the coexistence of heterogeneous wireless technologies. However, due to the physical layer incompatibility of WiFi and Bluetooth Low Energy (BLE), the researches about CTC between these two most wildly used wireless technologies are limited by now. In this paper, we propose WiBWi, a payload encoding-based bidirectional CTC scheme between BLE and WiFi, which can achieve near-optimal throughput and powerful robustness. For uplink, i.e., BLE to WiFi communication, WiBWi leverages a novel extended WiFi preamble detection rule and probabilistic inference based encode mapping to achieve fast and reliable communication. For downlink, i.e., WiFi to BLE communication, WiBWi introduces an encoding mapping scheme in the sight of BLE receiver with little modification to accomplish high throughput and robustness. Extensive evaluation shows that WiBWi can offer near-optimal throughput (near the maximum throughput of BLE) and extremely low bit error rate (less than 1%).
{"title":"WiBWi: Encoding-based Bidirectional Physical-Layer Cross-Technology Communication between BLE and WiFi","authors":"Yuanhe Shu, Jingwei Wang, L. Kong, Jiadi Yu, Guisong Yang, Yueping Cai, Zhen Wang, M. K. Khan","doi":"10.1109/ICPADS53394.2021.00050","DOIUrl":"https://doi.org/10.1109/ICPADS53394.2021.00050","url":null,"abstract":"The booming of mobile technologies and Internet of Things (IoTs) have facilitated the explosion of wireless devices and brought convenience to people's daily lives. Coming with the explosive growth of wireless devices, incompatibility of heterogeneous wireless technologies hindered the growing demands for everything connected. And spectrum sharing among heterogeneous wireless technologies has led to severe Cross-Technology Interference (CTI), which is a vital obstacle for network reliability and spectrum utilization. Researches in recent years have shown that Cross-Technology Communication (CTC) turns out to be a promising solution with broad perspective for the coexistence of heterogeneous wireless technologies. However, due to the physical layer incompatibility of WiFi and Bluetooth Low Energy (BLE), the researches about CTC between these two most wildly used wireless technologies are limited by now. In this paper, we propose WiBWi, a payload encoding-based bidirectional CTC scheme between BLE and WiFi, which can achieve near-optimal throughput and powerful robustness. For uplink, i.e., BLE to WiFi communication, WiBWi leverages a novel extended WiFi preamble detection rule and probabilistic inference based encode mapping to achieve fast and reliable communication. For downlink, i.e., WiFi to BLE communication, WiBWi introduces an encoding mapping scheme in the sight of BLE receiver with little modification to accomplish high throughput and robustness. Extensive evaluation shows that WiBWi can offer near-optimal throughput (near the maximum throughput of BLE) and extremely low bit error rate (less than 1%).","PeriodicalId":309508,"journal":{"name":"2021 IEEE 27th International Conference on Parallel and Distributed Systems (ICPADS)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133625889","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Accurate traffic prediction is of great importance in Intelligent Transportation System. This problem is very challenging due to the complex spatial and long-range temporal dependencies. Existing models generally suffer two limitations: (1) GCN-based methods usually use a fixed Laplacian matrix to model spatial dependencies, without considering their dynamics; (2) RNN and its variants are only capable of modeling a limited-range temporal dependencies, resulting in significant information loss. In this paper, we propose a novel spatial-temporal graph neural network (STNN), an end-to-end solution for traffic prediction that simultaneously captures dynamic spatial and long-range temporal dependencies. Specifically, STNN first uses a spatial attention network to model complex and dynamic spatial correlations, without any expensive matrix operations or relying on predefined road network topologies. Second, a temporal transformer network is utilized to model long-range temporal dependencies across multiple time steps, which considers not only the recent segment, but also the periodic dependencies of historical data. Making full use of historical data can alleviate the difficulty of obtaining real-time data and improve the prediction accuracy. Experiments are conducted on two real-world traffic datasets, and the results verify the effectiveness of the proposed model, especially in long-term traffic prediction.
{"title":"STNN: A Spatial-Temporal Graph Neural Network for Traffic Prediction","authors":"Xueyan Yin, Fei Li, Genze Wu, Pengfei Wang, Yanming Shen, Heng Qi, Baocai Yin","doi":"10.1109/ICPADS53394.2021.00024","DOIUrl":"https://doi.org/10.1109/ICPADS53394.2021.00024","url":null,"abstract":"Accurate traffic prediction is of great importance in Intelligent Transportation System. This problem is very challenging due to the complex spatial and long-range temporal dependencies. Existing models generally suffer two limitations: (1) GCN-based methods usually use a fixed Laplacian matrix to model spatial dependencies, without considering their dynamics; (2) RNN and its variants are only capable of modeling a limited-range temporal dependencies, resulting in significant information loss. In this paper, we propose a novel spatial-temporal graph neural network (STNN), an end-to-end solution for traffic prediction that simultaneously captures dynamic spatial and long-range temporal dependencies. Specifically, STNN first uses a spatial attention network to model complex and dynamic spatial correlations, without any expensive matrix operations or relying on predefined road network topologies. Second, a temporal transformer network is utilized to model long-range temporal dependencies across multiple time steps, which considers not only the recent segment, but also the periodic dependencies of historical data. Making full use of historical data can alleviate the difficulty of obtaining real-time data and improve the prediction accuracy. Experiments are conducted on two real-world traffic datasets, and the results verify the effectiveness of the proposed model, especially in long-term traffic prediction.","PeriodicalId":309508,"journal":{"name":"2021 IEEE 27th International Conference on Parallel and Distributed Systems (ICPADS)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117234571","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Malware detection has attracted widespread attention due to the growing malware sophistication. Machine learning based methods have been proposed to find traces of malware by analyzing network traffic. However, network traffic exhibits a series of growing and changing states, which makes it challenging to design a detection model that can detect malicious traffic over a long period without the need for costly retraining. In this paper, we present, IEdroid, an Android malicious network behavior detection method that leverages incremental ensembles for model update. Specifically, we train multiple classifiers to form an interim ensemble in distributed cluster environment, and update the interim ensemble by removing and adding classifiers. The generated model is composed of multiple interim ensembles that can adapt to the network traffic. We evaluated the performance of IEdroid using a dataset consisting of 98,565 benign and 41,267 malicious flows. Results show that IEdroid can effectively detect malicious traffic compared with state-of-the-art detection models. The experiment trained IEdroid on datasets incrementally for 10 times without a significant loss on accuracy, precision, recall, and F-Measure, compared with re-training from scratch with full data.
{"title":"IEdroid:Detecting Malicious Android Network Behavior Using Incremental Ensemble of Ensembles","authors":"Cong Liu, Anli Yan, Zhenxiang Chen, Haibo Zhang, Qiben Yan, Lizhi Peng, Chuan Zhao","doi":"10.1109/ICPADS53394.2021.00104","DOIUrl":"https://doi.org/10.1109/ICPADS53394.2021.00104","url":null,"abstract":"Malware detection has attracted widespread attention due to the growing malware sophistication. Machine learning based methods have been proposed to find traces of malware by analyzing network traffic. However, network traffic exhibits a series of growing and changing states, which makes it challenging to design a detection model that can detect malicious traffic over a long period without the need for costly retraining. In this paper, we present, IEdroid, an Android malicious network behavior detection method that leverages incremental ensembles for model update. Specifically, we train multiple classifiers to form an interim ensemble in distributed cluster environment, and update the interim ensemble by removing and adding classifiers. The generated model is composed of multiple interim ensembles that can adapt to the network traffic. We evaluated the performance of IEdroid using a dataset consisting of 98,565 benign and 41,267 malicious flows. Results show that IEdroid can effectively detect malicious traffic compared with state-of-the-art detection models. The experiment trained IEdroid on datasets incrementally for 10 times without a significant loss on accuracy, precision, recall, and F-Measure, compared with re-training from scratch with full data.","PeriodicalId":309508,"journal":{"name":"2021 IEEE 27th International Conference on Parallel and Distributed Systems (ICPADS)","volume":"62 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116973050","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-01DOI: 10.1109/ICPADS53394.2021.00058
Jianhua Gao, Weixing Ji, Jie Liu, Senhao Shao, Yizhuo Wang, Feng Shi
SpMV is a cost-dominant operation used in many iterative methods for solving large-scale sparse linear systems. However, irregular memory access of SpMV to the multiplied vector leads to low data locality and then harms the performance. This paper presents an adaptive multi-row folding of CSR (AMF-CSR) format for SpMV calculation on GPU. This new storage format supports the folding of the variable number of rows in order to achieve better load balancing in computation. AMF-CSR not only increases the density of non-zero elements in a folded row, thereby improving the access locality of the multiplied vector, but also merges an approximately equal number of nonzero elements in a folded row, hence achieving load balancing. The performance evaluation using 28 sparse matrices shows that the proposed SpMV algorithm based on AMF-CSR achieves the highest speedup of 4.11x and 3.62x on GTX 1080 Ti and Tesla V100 respectively against a fixed multi-row folding-based SpMV algorithm. Evaluation results using 450 regular sparse matrices and 450 irregular sparse matrices also show that AMF-CSR is superior to other SpMV implementations.
{"title":"AMF-CSR: Adaptive Multi-Row Folding of CSR for SpMV on GPU","authors":"Jianhua Gao, Weixing Ji, Jie Liu, Senhao Shao, Yizhuo Wang, Feng Shi","doi":"10.1109/ICPADS53394.2021.00058","DOIUrl":"https://doi.org/10.1109/ICPADS53394.2021.00058","url":null,"abstract":"SpMV is a cost-dominant operation used in many iterative methods for solving large-scale sparse linear systems. However, irregular memory access of SpMV to the multiplied vector leads to low data locality and then harms the performance. This paper presents an adaptive multi-row folding of CSR (AMF-CSR) format for SpMV calculation on GPU. This new storage format supports the folding of the variable number of rows in order to achieve better load balancing in computation. AMF-CSR not only increases the density of non-zero elements in a folded row, thereby improving the access locality of the multiplied vector, but also merges an approximately equal number of nonzero elements in a folded row, hence achieving load balancing. The performance evaluation using 28 sparse matrices shows that the proposed SpMV algorithm based on AMF-CSR achieves the highest speedup of 4.11x and 3.62x on GTX 1080 Ti and Tesla V100 respectively against a fixed multi-row folding-based SpMV algorithm. Evaluation results using 450 regular sparse matrices and 450 irregular sparse matrices also show that AMF-CSR is superior to other SpMV implementations.","PeriodicalId":309508,"journal":{"name":"2021 IEEE 27th International Conference on Parallel and Distributed Systems (ICPADS)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115543914","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-01DOI: 10.1109/ICPADS53394.2021.00075
Aibo Song, Bowen Peng, Jingyi Qiu, Yingying Xue, Mingyang Du
As a memory-based distributed big data computing framework, Spark has been widely used in big data processing systems. However, during the execution of Spark, due to the imbalance of input data distribution and the shortage of existing data partitioners in Spark, it is easy to cause partition skew problem and reduce the execution efficiency of Spark. Aiming at this problem, this paper proposes a balanced Spark data partitioner called BSDP (Balanced Spark Data Partitioner). By deeply analyzing the partitioning characteristics of Shuffle intermediate data, the Spark Shuffle intermediate data equalization partitioning model is established. The model aims to minimize the partition skew and find a Shuffle intermediate data equalization partitioning strategy. Based on the model, this paper designs and implements a data equalization partitioning algorithm of BSDP. This algorithm transforms the Shuffle intermediate data equalization partitioning problem into a classic List-Scheduling task scheduling problem, effectively realizes the balanced partitioning of Shuffle intermediate data. The experiment verifies that the BSDP can effectively realize the balanced partitioning of the Shuffle intermediate data and improve the execution efficiency of Spark.
Spark作为一种基于内存的分布式大数据计算框架，在大数据处理系统中得到了广泛的应用。然而，在Spark执行过程中，由于输入数据分布的不平衡以及Spark中现有数据分区的不足，容易造成分区倾斜问题，降低Spark的执行效率。针对这一问题，本文提出了一种平衡的Spark数据分区器BSDP (balanced Spark data partitioner)。通过深入分析Shuffle中间数据的分区特点，建立了Spark Shuffle中间数据均衡分区模型。该模型旨在最小化分区倾斜，并找到一种Shuffle中间数据均衡分区策略。基于该模型，设计并实现了一种BSDP数据均衡分区算法。该算法将Shuffle中间数据均衡分区问题转化为经典的List-Scheduling任务调度问题，有效地实现了Shuffle中间数据均衡分区。实验验证了BSDP可以有效地实现Shuffle中间数据的均衡分区，提高Spark的执行效率。
{"title":"BSDP: A Novel Balanced Spark Data Partitioner","authors":"Aibo Song, Bowen Peng, Jingyi Qiu, Yingying Xue, Mingyang Du","doi":"10.1109/ICPADS53394.2021.00075","DOIUrl":"https://doi.org/10.1109/ICPADS53394.2021.00075","url":null,"abstract":"As a memory-based distributed big data computing framework, Spark has been widely used in big data processing systems. However, during the execution of Spark, due to the imbalance of input data distribution and the shortage of existing data partitioners in Spark, it is easy to cause partition skew problem and reduce the execution efficiency of Spark. Aiming at this problem, this paper proposes a balanced Spark data partitioner called BSDP (Balanced Spark Data Partitioner). By deeply analyzing the partitioning characteristics of Shuffle intermediate data, the Spark Shuffle intermediate data equalization partitioning model is established. The model aims to minimize the partition skew and find a Shuffle intermediate data equalization partitioning strategy. Based on the model, this paper designs and implements a data equalization partitioning algorithm of BSDP. This algorithm transforms the Shuffle intermediate data equalization partitioning problem into a classic List-Scheduling task scheduling problem, effectively realizes the balanced partitioning of Shuffle intermediate data. The experiment verifies that the BSDP can effectively realize the balanced partitioning of the Shuffle intermediate data and improve the execution efficiency of Spark.","PeriodicalId":309508,"journal":{"name":"2021 IEEE 27th International Conference on Parallel and Distributed Systems (ICPADS)","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129737907","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-01DOI: 10.1109/ICPADS53394.2021.00108
Iryanto Jaya, Yusen Li, Wentong Cai
Cloud gaming abstracts the concept of traditional gaming and places the gaming activities on remote rendering servers (RSes). Although this allows heterogeneous devices to gain access to multiple game titles, latency issue is always unavoidable. Each game input must go through a complete round trip between the player's device and the cloud gaming server. Hence, cloud games are not as responsive as traditional computer games where the game logic runs locally. Moreover, in order to have an acceptable level of game playability, the latency level must be within a certain threshold. This also prevents some players who are located in remote regions from playing the game due to high latency. Therefore, in this paper, we employ edge servers in order to reach those players by activating lower capability RSes which are more geographically distributed. Furthermore, we also allow workload splitting of foreground and background rendering between edge and cloud RSes to ease the burden of each individual RS with a trade-off between cost and latency constraints. From our experiments, our architecture and allocation scheme results in reduction of play request rejections for up to 28% compared to traditional cloud gaming approach.
{"title":"Minimizing Play Request Rejection through Workload Splitting in Edge-Cloud Gaming","authors":"Iryanto Jaya, Yusen Li, Wentong Cai","doi":"10.1109/ICPADS53394.2021.00108","DOIUrl":"https://doi.org/10.1109/ICPADS53394.2021.00108","url":null,"abstract":"Cloud gaming abstracts the concept of traditional gaming and places the gaming activities on remote rendering servers (RSes). Although this allows heterogeneous devices to gain access to multiple game titles, latency issue is always unavoidable. Each game input must go through a complete round trip between the player's device and the cloud gaming server. Hence, cloud games are not as responsive as traditional computer games where the game logic runs locally. Moreover, in order to have an acceptable level of game playability, the latency level must be within a certain threshold. This also prevents some players who are located in remote regions from playing the game due to high latency. Therefore, in this paper, we employ edge servers in order to reach those players by activating lower capability RSes which are more geographically distributed. Furthermore, we also allow workload splitting of foreground and background rendering between edge and cloud RSes to ease the burden of each individual RS with a trade-off between cost and latency constraints. From our experiments, our architecture and allocation scheme results in reduction of play request rejections for up to 28% compared to traditional cloud gaming approach.","PeriodicalId":309508,"journal":{"name":"2021 IEEE 27th International Conference on Parallel and Distributed Systems (ICPADS)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130301536","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：