Proceedings of the 5th ACM SIGMOD Joint International Workshop on Graph Data Management Experiences & Systems (GRADES) and Network Data Analytics (NDA)最新文献
In this paper, we present DynaGraph, a system that supports dynamic Graph Neural Networks (GNNs) efficiently. Based on the observation that existing proposals for dynamic GNN architectures combine techniques for structural and temporal information encoding independently, DynaGraph proposes novel techniques that enable cross optimizations across these tasks. It uses cached message passing and timestep fusion to significantly reduce the overhead associated with dynamic GNN processing. It further proposes a simple distributed data-parallel dynamic graph processing strategy that enables scalable dynamic GNN computation. Our evaluation of DynaGraph on a variety of dynamic GNN architectures and use cases shows a speedup of up to 2.7X compared to existing state-of-the-art frameworks.
{"title":"DynaGraph","authors":"M. Guan, A. Iyer, Taesoo Kim","doi":"10.1145/3534540.3534691","DOIUrl":"https://doi.org/10.1145/3534540.3534691","url":null,"abstract":"In this paper, we present DynaGraph, a system that supports dynamic Graph Neural Networks (GNNs) efficiently. Based on the observation that existing proposals for dynamic GNN architectures combine techniques for structural and temporal information encoding independently, DynaGraph proposes novel techniques that enable cross optimizations across these tasks. It uses cached message passing and timestep fusion to significantly reduce the overhead associated with dynamic GNN processing. It further proposes a simple distributed data-parallel dynamic graph processing strategy that enables scalable dynamic GNN computation. Our evaluation of DynaGraph on a variety of dynamic GNN architectures and use cases shows a speedup of up to 2.7X compared to existing state-of-the-art frameworks.","PeriodicalId":406863,"journal":{"name":"Proceedings of the 5th ACM SIGMOD Joint International Workshop on Graph Data Management Experiences & Systems (GRADES) and Network Data Analytics (NDA)","volume":"75 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124743719","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}
This paper demonstrates ShaderNet --- our graph analytics framework with shader codes, which are machine-level codes and are important for GPU designers to tune the hardware, e.g., adjusting clock speeds and voltages. Due to a wide spectrum of use-cases of modern GPUs, engineers generally find it difficult to manually inspect a large number of shader codes emerging from these applications. To this end, we present a system, ShaderNet, which converts shader codes into graphs, and applies advanced graph mining and machine learning techniques to simplify shader graphs analysis in an effective and explainable manner. By studying shader codes' evolution with temporal graphs analysis and structure mining with frequent subgraphs, we demonstrate several key functionalities of our framework, such as a frame's scene detection, clustering scenes, and a new application's inefficient shaders prediction, which can accelerate GPU's performance tuning. Our code base and demonstration video are at: https://lzlz15.github.io/D_E_M_O/.
{"title":"ShaderNet","authors":"Lin Zhao, Arijit Khan, Robby Luo","doi":"10.1145/3534540.3534688","DOIUrl":"https://doi.org/10.1145/3534540.3534688","url":null,"abstract":"This paper demonstrates ShaderNet --- our graph analytics framework with shader codes, which are machine-level codes and are important for GPU designers to tune the hardware, e.g., adjusting clock speeds and voltages. Due to a wide spectrum of use-cases of modern GPUs, engineers generally find it difficult to manually inspect a large number of shader codes emerging from these applications. To this end, we present a system, ShaderNet, which converts shader codes into graphs, and applies advanced graph mining and machine learning techniques to simplify shader graphs analysis in an effective and explainable manner. By studying shader codes' evolution with temporal graphs analysis and structure mining with frequent subgraphs, we demonstrate several key functionalities of our framework, such as a frame's scene detection, clustering scenes, and a new application's inefficient shaders prediction, which can accelerate GPU's performance tuning. Our code base and demonstration video are at: https://lzlz15.github.io/D_E_M_O/.","PeriodicalId":406863,"journal":{"name":"Proceedings of the 5th ACM SIGMOD Joint International Workshop on Graph Data Management Experiences & Systems (GRADES) and Network Data Analytics (NDA)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115458567","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}
Andrew McCrabb, H. Nigatu, Absalat Getachew, V. Bertacco
Dynamic graph processing, execution on vertex-edge graphs that change over time, is quickly becoming a key computing need of the twenty-first century. Dynamic graph algorithms unlock real-time optimization solutions and a wide range of data-mining applications in logistics, finance, marketing, healthcare, and social media, among many others. However, graph algorithms are extremely memory-bound (i.e., their performance is limited by the bandwidth of memory accesses on the underlying hardware platform, rather than the compute capacity). Moreover, dynamic graph algorithms are being applied to increasingly-large datasets, further straining the memory systems and reducing performance. As a result, additional research is needed to leverage new memory technologies for faster, more efficient, dynamic graph-based processing. Such research is difficult without access to hitherto unavailable industrial-scale dynamic graph datasets to evaluate solutions. In this work, we present DyGraph, a dynamic graph synthetic dataset generator paired with a collection of real-world graphs in the domains of social media, recommendation systems, and fintech. We demonstrate the breadth of graph features represented in this repository and evaluate the DyGraph Generator's ability to generate synthetic graphs that mimic these real datasets. In our case study, we find that the degree distribution of DyGraph Generator datasets correlate 3 to 5.5 times more closely to real-world datasets than Power Law models, paving the way for much-needed research for high-performance dynamic graph processing.
{"title":"DyGraph","authors":"Andrew McCrabb, H. Nigatu, Absalat Getachew, V. Bertacco","doi":"10.1145/3534540.3534692","DOIUrl":"https://doi.org/10.1145/3534540.3534692","url":null,"abstract":"Dynamic graph processing, execution on vertex-edge graphs that change over time, is quickly becoming a key computing need of the twenty-first century. Dynamic graph algorithms unlock real-time optimization solutions and a wide range of data-mining applications in logistics, finance, marketing, healthcare, and social media, among many others. However, graph algorithms are extremely memory-bound (i.e., their performance is limited by the bandwidth of memory accesses on the underlying hardware platform, rather than the compute capacity). Moreover, dynamic graph algorithms are being applied to increasingly-large datasets, further straining the memory systems and reducing performance. As a result, additional research is needed to leverage new memory technologies for faster, more efficient, dynamic graph-based processing. Such research is difficult without access to hitherto unavailable industrial-scale dynamic graph datasets to evaluate solutions. In this work, we present DyGraph, a dynamic graph synthetic dataset generator paired with a collection of real-world graphs in the domains of social media, recommendation systems, and fintech. We demonstrate the breadth of graph features represented in this repository and evaluate the DyGraph Generator's ability to generate synthetic graphs that mimic these real datasets. In our case study, we find that the degree distribution of DyGraph Generator datasets correlate 3 to 5.5 times more closely to real-world datasets than Power Law models, paving the way for much-needed research for high-performance dynamic graph processing.","PeriodicalId":406863,"journal":{"name":"Proceedings of the 5th ACM SIGMOD Joint International Workshop on Graph Data Management Experiences & Systems (GRADES) and Network Data Analytics (NDA)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124365828","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}
Proceedings of the 5th ACM SIGMOD Joint International Workshop on Graph Data Management Experiences & Systems (GRADES) and Network Data Analytics (NDA)
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：