{"title":"Session details: Keynote 2","authors":"Jinoh Kim","doi":"10.1145/3240508.3286918","DOIUrl":"https://doi.org/10.1145/3240508.3286918","url":null,"abstract":"","PeriodicalId":339207,"journal":{"name":"Proceedings of the 2021 on Systems and Network Telemetry and Analytics","volume":"46 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127486626","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}
Jianwei Hao, Piyush Subedi, I. Kim, Lakshmish Ramaswamy
Significant advances in hardware capabilities and the availability of enormous data sets have led to the rise and penetration of artificial intelligence (AI) and deep learning (DL) in various domains. Considerable efforts have been put forth in academia and industry to make these computationally demanding DL tasks work on resource-constrained edge devices. However, performing DL tasks on edge devices is still challenging due to the diversity of DNN (Deep Neural Networks) architectures and heterogeneity of edge devices. This study evaluates and characterizes the performance and resource heterogeneity in various edge devices for performing DL tasks. We benchmark various DNN models for image classification on a set of edge devices ranging from the widely popular and relatively less powerful Raspberry Pi to GPU-equipped high-performance edge devices like Jetson Xavier NX. We also compare and contrast the performance of three widely-used DL frameworks when used in these edge devices. We report DL inference throughput, CPU and memory usage, power consumption, and frameworks' initialization overhead, which are the most critical factors for characterizing DL tasks on edge devices. Additionally, we provide our insights and findings, which will provide a better idea of how compatible or feasible edge devices are for running DL applications.
{"title":"Characterizing Resource Heterogeneity in Edge Devices for Deep Learning Inferences","authors":"Jianwei Hao, Piyush Subedi, I. Kim, Lakshmish Ramaswamy","doi":"10.1145/3452411.3464446","DOIUrl":"https://doi.org/10.1145/3452411.3464446","url":null,"abstract":"Significant advances in hardware capabilities and the availability of enormous data sets have led to the rise and penetration of artificial intelligence (AI) and deep learning (DL) in various domains. Considerable efforts have been put forth in academia and industry to make these computationally demanding DL tasks work on resource-constrained edge devices. However, performing DL tasks on edge devices is still challenging due to the diversity of DNN (Deep Neural Networks) architectures and heterogeneity of edge devices. This study evaluates and characterizes the performance and resource heterogeneity in various edge devices for performing DL tasks. We benchmark various DNN models for image classification on a set of edge devices ranging from the widely popular and relatively less powerful Raspberry Pi to GPU-equipped high-performance edge devices like Jetson Xavier NX. We also compare and contrast the performance of three widely-used DL frameworks when used in these edge devices. We report DL inference throughput, CPU and memory usage, power consumption, and frameworks' initialization overhead, which are the most critical factors for characterizing DL tasks on edge devices. Additionally, we provide our insights and findings, which will provide a better idea of how compatible or feasible edge devices are for running DL applications.","PeriodicalId":339207,"journal":{"name":"Proceedings of the 2021 on Systems and Network Telemetry and Analytics","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127086203","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}
{"title":"Session details: Technical Session 1","authors":"M. Cafaro","doi":"10.1145/3470766","DOIUrl":"https://doi.org/10.1145/3470766","url":null,"abstract":"","PeriodicalId":339207,"journal":{"name":"Proceedings of the 2021 on Systems and Network Telemetry and Analytics","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127648060","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 article describes the keynote speech on INODE presented at Fourth International Workshop on Systems and Network Telemetry and Analytics (SNTA) which is collocated with International ACM Symposium on High-Performance Parallel and Distributed Computing (HPDC) on June 21 in Stockholm, Sweden.
{"title":"INODE - Intelligence Open Data Exploration","authors":"Kurt Stockinger","doi":"10.1145/3452411.3464448","DOIUrl":"https://doi.org/10.1145/3452411.3464448","url":null,"abstract":"This article describes the keynote speech on INODE presented at Fourth International Workshop on Systems and Network Telemetry and Analytics (SNTA) which is collocated with International ACM Symposium on High-Performance Parallel and Distributed Computing (HPDC) on June 21 in Stockholm, Sweden.","PeriodicalId":339207,"journal":{"name":"Proceedings of the 2021 on Systems and Network Telemetry and Analytics","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116721448","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}
The emergence of network function virtualization~(NFV) has revolutionized the infrastructure and service management of network architecture by reducing the cost and complexity of network service deployment. However, finding the optimal placement of virtual network functions (VNFs) is an NP-complete problem. Existing solutions base on either Integer Linear Programming~(ILP), or greedy algorithms. But, solving ILP can be time consuming and the approximation of greedy is not bounded. Hence, neither of them can make quick and accurate placement decisions, especially under dynamic traffic workload. Therefore, we propose a hybrid method that combines the two approaches together to achieve up to 45% service profit improvement with less computation time comparing to the traditional static ILP approach.
{"title":"A Hybrid Virtual Network Function Placement Strategy for Maximizing the Profit of Network Service Deployment over Dynamic Workload","authors":"Chi-Chen Yang, J. Chou","doi":"10.1145/3452411.3464440","DOIUrl":"https://doi.org/10.1145/3452411.3464440","url":null,"abstract":"The emergence of network function virtualization~(NFV) has revolutionized the infrastructure and service management of network architecture by reducing the cost and complexity of network service deployment. However, finding the optimal placement of virtual network functions (VNFs) is an NP-complete problem. Existing solutions base on either Integer Linear Programming~(ILP), or greedy algorithms. But, solving ILP can be time consuming and the approximation of greedy is not bounded. Hence, neither of them can make quick and accurate placement decisions, especially under dynamic traffic workload. Therefore, we propose a hybrid method that combines the two approaches together to achieve up to 45% service profit improvement with less computation time comparing to the traditional static ILP approach.","PeriodicalId":339207,"journal":{"name":"Proceedings of the 2021 on Systems and Network Telemetry and Analytics","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133403266","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}
Elizabeth Copps, Huiyi Zhang, A. Sim, Kesheng Wu, I. Monga, C. Guok, F. Würthwein, Diego Davila, E. Hernandez
The volume of data moving through a network increases with new scientific experiments and simulations. Network bandwidth requirements also increase proportionally to deliver data within a certain time frame. We observe that a significant portion of the popular dataset is transferred multiple times to different users as well as to the same user for various reasons. In-network data caching for the shared data has shown to reduce the redundant data transfers and consequently save network traffic volume. In addition, overall application performance is expected to improve with in-network caching because access to the locally cached data results in lower latency. This paper shows how much data was shared over the study period, how much network traffic volume was consequently saved, and how much the temporary in-network caching increased the scientific application performance. It also analyzes data access patterns in applications and the impacts of caching nodes on the regional data repository. From the results, we observed that the network bandwidth demand was reduced by nearly a factor of 3 over the study period.
{"title":"Analyzing Scientific Data Sharing Patterns for In-network Data Caching","authors":"Elizabeth Copps, Huiyi Zhang, A. Sim, Kesheng Wu, I. Monga, C. Guok, F. Würthwein, Diego Davila, E. Hernandez","doi":"10.1145/3452411.3464441","DOIUrl":"https://doi.org/10.1145/3452411.3464441","url":null,"abstract":"The volume of data moving through a network increases with new scientific experiments and simulations. Network bandwidth requirements also increase proportionally to deliver data within a certain time frame. We observe that a significant portion of the popular dataset is transferred multiple times to different users as well as to the same user for various reasons. In-network data caching for the shared data has shown to reduce the redundant data transfers and consequently save network traffic volume. In addition, overall application performance is expected to improve with in-network caching because access to the locally cached data results in lower latency. This paper shows how much data was shared over the study period, how much network traffic volume was consequently saved, and how much the temporary in-network caching increased the scientific application performance. It also analyzes data access patterns in applications and the impacts of caching nodes on the regional data repository. From the results, we observed that the network bandwidth demand was reduced by nearly a factor of 3 over the study period.","PeriodicalId":339207,"journal":{"name":"Proceedings of the 2021 on Systems and Network Telemetry and Analytics","volume":"56 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121438033","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}
{"title":"Session details: Technical Session 2","authors":"Jinoh Kim","doi":"10.1145/3470768","DOIUrl":"https://doi.org/10.1145/3470768","url":null,"abstract":"","PeriodicalId":339207,"journal":{"name":"Proceedings of the 2021 on Systems and Network Telemetry and Analytics","volume":"2020 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117058631","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}
Today's enterprise networks has revolutionized by the emerging technology called Network function virtualization (NFV), which is a type of data center network architecture proposed by the European Telecommunications Standards Institute (ETSI). NFV uses virtualization techniques to implement various Network Functions (NFs) like firewall, load balancer etc. from dedicated network devices to virtualized instances in commodity servers. This virtualized instance is called as Virtual Network Function (VNF). The purpose of VNF is to process NFs in order to accomplish a specific task. Traditionally these NFs were implemented on dedicated network equipment called middleboxes. Although these middleboxes are capable of processing heavy workloads, they are expensive, inflexible and require experts to maintain them. Therefore, NFV has the potential to substitute these middleboxes with virtualized instances in cloud datacenters, and hence greatly reduces the Operational Expenditure (OPEX) and Capital Expenditure (CAPEX) of networks by making it cheaper, flexible and scalable. This talk will share the recent advances and future challenges on how to build the infrastructure for hosting and managing NFVs. In particularly, we will focus on two of the most important topics in this research direction. First is the VNF placement problem, which aims to find the best mapping decision between VNF instances and physical resources. It has significant impact to the network operation cost, and service quality, but it is also known to be a NP-hard problem. So the problem has been actively studied by the research community. The second topic is Cloud-native/Container Network Function (CNF), which aims to minimize the overhead of traditional virtualization technique for network function using container-based technologies. Hence, it has drawn growing interests from industry to build the NFV infrastructure for CNF, but many new challenges remain to be addressed and studied.
{"title":"Recent Advances and Future Challenges for Network Function Virtualization Infrastructure","authors":"J. Chou","doi":"10.1145/3452411.3464449","DOIUrl":"https://doi.org/10.1145/3452411.3464449","url":null,"abstract":"Today's enterprise networks has revolutionized by the emerging technology called Network function virtualization (NFV), which is a type of data center network architecture proposed by the European Telecommunications Standards Institute (ETSI). NFV uses virtualization techniques to implement various Network Functions (NFs) like firewall, load balancer etc. from dedicated network devices to virtualized instances in commodity servers. This virtualized instance is called as Virtual Network Function (VNF). The purpose of VNF is to process NFs in order to accomplish a specific task. Traditionally these NFs were implemented on dedicated network equipment called middleboxes. Although these middleboxes are capable of processing heavy workloads, they are expensive, inflexible and require experts to maintain them. Therefore, NFV has the potential to substitute these middleboxes with virtualized instances in cloud datacenters, and hence greatly reduces the Operational Expenditure (OPEX) and Capital Expenditure (CAPEX) of networks by making it cheaper, flexible and scalable. This talk will share the recent advances and future challenges on how to build the infrastructure for hosting and managing NFVs. In particularly, we will focus on two of the most important topics in this research direction. First is the VNF placement problem, which aims to find the best mapping decision between VNF instances and physical resources. It has significant impact to the network operation cost, and service quality, but it is also known to be a NP-hard problem. So the problem has been actively studied by the research community. The second topic is Cloud-native/Container Network Function (CNF), which aims to minimize the overhead of traditional virtualization technique for network function using container-based technologies. Hence, it has drawn growing interests from industry to build the NFV infrastructure for CNF, but many new challenges remain to be addressed and studied.","PeriodicalId":339207,"journal":{"name":"Proceedings of the 2021 on Systems and Network Telemetry and Analytics","volume":"90 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130134925","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}
Automated network intrusion detection systems (NIDS) continuously monitor the network traffic to detect attacks or/and anomalies. These systems need to be able to detect attacks and alert network engineers in real-time. Therefore, modern NIDS are built using complex machine learning algorithms that require large training datasets and are time-consuming to train. The proposed work shows that machine learning algorithms from the RAPIDS cuML library on Graphics Processing Units (GPUs) can speed-up the training process on large scale datasets. This approach is able to reduce the training time while providing high accuracy and performance. We demonstrate the proposed approach on a large subset of data extracted from the Aegean Wi-Fi Intrusion Dataset (AWID). Multiple classification experiments were performed on both CPU and GPU. We achieve up to 65x acceleration of training several machine learning methods by moving most of the pipeline computations to the GPU and leveraging the new cuML library as well as the GPU version of the CatBoost library.
{"title":"GPU-based Classification for Wireless Intrusion Detection","authors":"A. Lazar, A. Sim, Kesheng Wu","doi":"10.1145/3452411.3464445","DOIUrl":"https://doi.org/10.1145/3452411.3464445","url":null,"abstract":"Automated network intrusion detection systems (NIDS) continuously monitor the network traffic to detect attacks or/and anomalies. These systems need to be able to detect attacks and alert network engineers in real-time. Therefore, modern NIDS are built using complex machine learning algorithms that require large training datasets and are time-consuming to train. The proposed work shows that machine learning algorithms from the RAPIDS cuML library on Graphics Processing Units (GPUs) can speed-up the training process on large scale datasets. This approach is able to reduce the training time while providing high accuracy and performance. We demonstrate the proposed approach on a large subset of data extracted from the Aegean Wi-Fi Intrusion Dataset (AWID). Multiple classification experiments were performed on both CPU and GPU. We achieve up to 65x acceleration of training several machine learning methods by moving most of the pipeline computations to the GPU and leveraging the new cuML library as well as the GPU version of the CatBoost library.","PeriodicalId":339207,"journal":{"name":"Proceedings of the 2021 on Systems and Network Telemetry and Analytics","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130941321","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}
Yumeng Wang, Kesheng Wu, A. Sim, Shinjae Yoo, S. Misawa
Large scientific projects are increasing relying on analyses of data for their new discoveries; and a number of different data management systems have been developed to serve this scientific projects. In the work-in-progress paper, we describe an effort on understanding the data access patterns of one of these data management systems, dCache. This particular deployment of dCache acts as a disk cache in front of a large tape storage system primarily containing high-energy physics data. Based on the 15-month dCache logs, the cache is only accessing the tape system once for over 50 file requests, which indicates that it is effective as a disk cache. The on-disk files are repeated used, more than three times a day. We have also identified a number of unusual access patterns that are worth further investigation.
{"title":"Access Patterns to Disk Cache for Large Scientific Archive","authors":"Yumeng Wang, Kesheng Wu, A. Sim, Shinjae Yoo, S. Misawa","doi":"10.1145/3452411.3464444","DOIUrl":"https://doi.org/10.1145/3452411.3464444","url":null,"abstract":"Large scientific projects are increasing relying on analyses of data for their new discoveries; and a number of different data management systems have been developed to serve this scientific projects. In the work-in-progress paper, we describe an effort on understanding the data access patterns of one of these data management systems, dCache. This particular deployment of dCache acts as a disk cache in front of a large tape storage system primarily containing high-energy physics data. Based on the 15-month dCache logs, the cache is only accessing the tape system once for over 50 file requests, which indicates that it is effective as a disk cache. The on-disk files are repeated used, more than three times a day. We have also identified a number of unusual access patterns that are worth further investigation.","PeriodicalId":339207,"journal":{"name":"Proceedings of the 2021 on Systems and Network Telemetry and Analytics","volume":"62 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125052396","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：