Pub Date : 2022-06-27DOI: 10.1109/NetSoft54395.2022.9844038
L. Gifre, C. Manso, R. Casellas, R. Martínez, R. Vilalta, R. Muñoz
This demonstration will showcase the end-to-end orchestration of virtual network functions in the full-fledged ADRENALINE Testbed Cloud Platform expanding from the edge to the cloud. The Management and Orchestration (MANO) software ETSI OpenSource MANO (OSM) is used to deploy and handle a multi-site network service involving both edge and core Data Centers (DCs). Besides, the inter- and intra-DC connectivity is directly managed by a novel OSM WAN Infrastructure Manager (WIM) connector using the Transport API (TAPI) interface, thus completely abstracting the details of the underlying SDN controllers handling the programmability of the WAN network interconnecting the DCs.
此演示将展示在成熟的ADRENALINE Testbed云平台中从边缘扩展到云的虚拟网络功能的端到端编排。管理和业务流程(MANO)软件ETSI OpenSource MANO (OSM)用于部署和处理涉及边缘和核心数据中心(dc)的多站点网络服务。此外，数据中心间和数据中心内的连接由使用传输API (TAPI)接口的新型OSM广域网基础设施管理器(WIM)连接器直接管理，从而完全抽象了处理连接数据中心的广域网可编程性的底层SDN控制器的细节。
{"title":"Experimental Demonstration of End-to-end NFV Orchestration on Top of the ADRENALINE Testbed","authors":"L. Gifre, C. Manso, R. Casellas, R. Martínez, R. Vilalta, R. Muñoz","doi":"10.1109/NetSoft54395.2022.9844038","DOIUrl":"https://doi.org/10.1109/NetSoft54395.2022.9844038","url":null,"abstract":"This demonstration will showcase the end-to-end orchestration of virtual network functions in the full-fledged ADRENALINE Testbed Cloud Platform expanding from the edge to the cloud. The Management and Orchestration (MANO) software ETSI OpenSource MANO (OSM) is used to deploy and handle a multi-site network service involving both edge and core Data Centers (DCs). Besides, the inter- and intra-DC connectivity is directly managed by a novel OSM WAN Infrastructure Manager (WIM) connector using the Transport API (TAPI) interface, thus completely abstracting the details of the underlying SDN controllers handling the programmability of the WAN network interconnecting the DCs.","PeriodicalId":125799,"journal":{"name":"2022 IEEE 8th International Conference on Network Softwarization (NetSoft)","volume":"78 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124103201","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 : 2022-06-27DOI: 10.1109/NetSoft54395.2022.9844061
D. Borsatti, W. Cerroni, S. Clayman
The possibility of managing network infrastructures through software-based programmable interfaces is becoming a cornerstone in the evolution of communication networks. The Intent-Based Networking (IBN) paradigm is a novel declarative approach towards network management proposed by a few Standards Developing Organizations. This paradigm offers a high-level interface for network management that abstracts the underlying network infrastructure and allows the specification of network directives using natural language. Since the IBN concept is based on a declarative approach to network management and programmability, we argue that the use of declarative programming to achieve IBN could uncover valuable insights for this new network paradigm. This paper proposes a formalization of this declarative paradigm obtained with concepts from category theory. Taking this approach to Intent, an initial implementation of this formalization is presented using Haskell, a well-known functional programming language.
{"title":"From Category Theory to Functional Programming: A Formal Representation of Intent","authors":"D. Borsatti, W. Cerroni, S. Clayman","doi":"10.1109/NetSoft54395.2022.9844061","DOIUrl":"https://doi.org/10.1109/NetSoft54395.2022.9844061","url":null,"abstract":"The possibility of managing network infrastructures through software-based programmable interfaces is becoming a cornerstone in the evolution of communication networks. The Intent-Based Networking (IBN) paradigm is a novel declarative approach towards network management proposed by a few Standards Developing Organizations. This paradigm offers a high-level interface for network management that abstracts the underlying network infrastructure and allows the specification of network directives using natural language. Since the IBN concept is based on a declarative approach to network management and programmability, we argue that the use of declarative programming to achieve IBN could uncover valuable insights for this new network paradigm. This paper proposes a formalization of this declarative paradigm obtained with concepts from category theory. Taking this approach to Intent, an initial implementation of this formalization is presented using Haskell, a well-known functional programming language.","PeriodicalId":125799,"journal":{"name":"2022 IEEE 8th International Conference on Network Softwarization (NetSoft)","volume":"66 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124424867","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 : 2022-06-27DOI: 10.1109/NetSoft54395.2022.9844109
Sherry Bai, Hyojoon Kim, J. Rexford
Operating System (OS) fingerprinting allows network administrators to identify which operating systems are running on the hosts communicating over their network. This information is useful for detecting OS-specific vulnerabilities and for administering OS-related security policies that block, rate-limit, or redirect traffic. Passive fingerprinting can identify hosts’ OS types without active probes that introduce additional network load. However, existing software-based passive fingerprinting tools cannot keep up with the traffic in high-speed networks. This paper presents P40f, a tool that runs on programmable switch hardware to perform OS fingerprinting and apply security policies at line rate. Unlike p0f, P40f can fingerprint devices’ OS types and react to it (e.g., drop, rate-limit) in real time directly in the switch, without requiring any control-plane messages. P40f is a P4 implementation of an existing software tool, p0f. We present our prototype implemented with the P4 language, which compiles and runs on the Intel Tofino switch. We present experiments against packet traces from a real campus network, and make our code publicly available.
{"title":"Passive OS Fingerprinting on Commodity Switches","authors":"Sherry Bai, Hyojoon Kim, J. Rexford","doi":"10.1109/NetSoft54395.2022.9844109","DOIUrl":"https://doi.org/10.1109/NetSoft54395.2022.9844109","url":null,"abstract":"Operating System (OS) fingerprinting allows network administrators to identify which operating systems are running on the hosts communicating over their network. This information is useful for detecting OS-specific vulnerabilities and for administering OS-related security policies that block, rate-limit, or redirect traffic. Passive fingerprinting can identify hosts’ OS types without active probes that introduce additional network load. However, existing software-based passive fingerprinting tools cannot keep up with the traffic in high-speed networks. This paper presents P40f, a tool that runs on programmable switch hardware to perform OS fingerprinting and apply security policies at line rate. Unlike p0f, P40f can fingerprint devices’ OS types and react to it (e.g., drop, rate-limit) in real time directly in the switch, without requiring any control-plane messages. P40f is a P4 implementation of an existing software tool, p0f. We present our prototype implemented with the P4 language, which compiles and runs on the Intel Tofino switch. We present experiments against packet traces from a real campus network, and make our code publicly available.","PeriodicalId":125799,"journal":{"name":"2022 IEEE 8th International Conference on Network Softwarization (NetSoft)","volume":"48 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129073975","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 : 2022-06-27DOI: 10.1109/NetSoft54395.2022.9844096
Daniel Lukaszewski, G. Xie
Protocol customizations primarily come in two forms: those driven by public extensions to open standard protocols; and dialecting and performance tuning driven by an enterprise network’s private security and performance needs. Current deployment of protocol customizations is mostly ad hoc, through manual configuration or script programs that are highly specialized to each customization. The method lacks the agility necessary to support the relatively high tempo of private customizations. Also, it is common for today’s protocol customization efforts to experience middlebox interference. In this paper, we propose a systematic framework of network-wide orchestration and continuous management of protocol customization for enterprise and data-center networks. By introducing a logically centralized orchestrator along with a layer 4.5 fine-grained device customization solution, our framework will allow operators to deploy and monitor customized flows from a single vantage point, providing timely detection of rogue devices as well as real-time coordination of middlebox traversal. Results from prototyping and experimentation confirm utility of our framework and show that the framework incurs modest processing overhead, at the levels of 3% and 1% for sample customized flows and non-customized flows, respectively.
{"title":"Towards Software Defined Layer 4.5 Customization","authors":"Daniel Lukaszewski, G. Xie","doi":"10.1109/NetSoft54395.2022.9844096","DOIUrl":"https://doi.org/10.1109/NetSoft54395.2022.9844096","url":null,"abstract":"Protocol customizations primarily come in two forms: those driven by public extensions to open standard protocols; and dialecting and performance tuning driven by an enterprise network’s private security and performance needs. Current deployment of protocol customizations is mostly ad hoc, through manual configuration or script programs that are highly specialized to each customization. The method lacks the agility necessary to support the relatively high tempo of private customizations. Also, it is common for today’s protocol customization efforts to experience middlebox interference. In this paper, we propose a systematic framework of network-wide orchestration and continuous management of protocol customization for enterprise and data-center networks. By introducing a logically centralized orchestrator along with a layer 4.5 fine-grained device customization solution, our framework will allow operators to deploy and monitor customized flows from a single vantage point, providing timely detection of rogue devices as well as real-time coordination of middlebox traversal. Results from prototyping and experimentation confirm utility of our framework and show that the framework incurs modest processing overhead, at the levels of 3% and 1% for sample customized flows and non-customized flows, respectively.","PeriodicalId":125799,"journal":{"name":"2022 IEEE 8th International Conference on Network Softwarization (NetSoft)","volume":"150 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115892027","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 : 2022-06-27DOI: 10.1109/NetSoft54395.2022.9844118
Matthews Jose, Kahina Lazri, J. François, O. Festor
The current generation of networks empowers the use of programmable switches whose behaviour can be defined using languages like P4. Nevertheless, these languages do not support network-wide deployment of stateful real-value functions. This paper presents NetREC, an extension of RMT programmable data planes designed to enable stateful real-value functions computation across multiple switches. NetREC first decomposes the real-value functions into a dependency graph of elementary operations that are distributed among the network. This distribution is carried out by dynamically generating and solving an integer linear program. We deploy a prototype of NetREC on a network of Tofino switches and demonstrate its capability of computing recursive real-value functions like exponential weighted moving average.
{"title":"NetREC: Network-wide in-network REal-value Computation","authors":"Matthews Jose, Kahina Lazri, J. François, O. Festor","doi":"10.1109/NetSoft54395.2022.9844118","DOIUrl":"https://doi.org/10.1109/NetSoft54395.2022.9844118","url":null,"abstract":"The current generation of networks empowers the use of programmable switches whose behaviour can be defined using languages like P4. Nevertheless, these languages do not support network-wide deployment of stateful real-value functions. This paper presents NetREC, an extension of RMT programmable data planes designed to enable stateful real-value functions computation across multiple switches. NetREC first decomposes the real-value functions into a dependency graph of elementary operations that are distributed among the network. This distribution is carried out by dynamically generating and solving an integer linear program. We deploy a prototype of NetREC on a network of Tofino switches and demonstrate its capability of computing recursive real-value functions like exponential weighted moving average.","PeriodicalId":125799,"journal":{"name":"2022 IEEE 8th International Conference on Network Softwarization (NetSoft)","volume":"50 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115165669","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 : 2022-06-27DOI: 10.1109/NetSoft54395.2022.9844053
Badre Bousalem, Vinicius F. Silva, R. Langar, Sylvain Cherrier
In this demo, we present a 5G prototype for attacks detection and mitigation in sliced networks leveraging Machine Learning (ML). Our prototype, based on OpenAirInterface, allows creating network slices on demand and managing physical resources dynamically according to the users’ behavior, while considering the inputs from a northbound Software Defined Network (SDN) application. We focus here on Distributed Denial of Service (DDoS) attacks, where one or multiple malicious users generate attacks on the 5G Core Network. Based on our developed ML module, we show that our prototype is able to detect such attacks, then automatically creates a sinkhole-type slice with a small portion of physical resources, and isolates the malicious users within this slice to mitigate the attackers’ action. We demonstrate the effectiveness of our approach by showing the decrease in the network throughput for the malicious users by a factor of 15, while maintaining a high network throughput for benign users.
{"title":"Deep Learning-based Approach for DDoS Attacks Detection and Mitigation in 5G and Beyond Mobile Networks","authors":"Badre Bousalem, Vinicius F. Silva, R. Langar, Sylvain Cherrier","doi":"10.1109/NetSoft54395.2022.9844053","DOIUrl":"https://doi.org/10.1109/NetSoft54395.2022.9844053","url":null,"abstract":"In this demo, we present a 5G prototype for attacks detection and mitigation in sliced networks leveraging Machine Learning (ML). Our prototype, based on OpenAirInterface, allows creating network slices on demand and managing physical resources dynamically according to the users’ behavior, while considering the inputs from a northbound Software Defined Network (SDN) application. We focus here on Distributed Denial of Service (DDoS) attacks, where one or multiple malicious users generate attacks on the 5G Core Network. Based on our developed ML module, we show that our prototype is able to detect such attacks, then automatically creates a sinkhole-type slice with a small portion of physical resources, and isolates the malicious users within this slice to mitigate the attackers’ action. We demonstrate the effectiveness of our approach by showing the decrease in the network throughput for the malicious users by a factor of 15, while maintaining a high network throughput for benign users.","PeriodicalId":125799,"journal":{"name":"2022 IEEE 8th International Conference on Network Softwarization (NetSoft)","volume":"55 9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127221375","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 : 2022-06-27DOI: 10.1109/NetSoft54395.2022.9844042
Oscar Delgado, B. Jaumard, Zhiyi Ding, Fadi Bishay, Vincent Bissonnette
Current 5G and beyond research explores how to leverage network virtualization, which allows network operators to partition the network into multiple independent slices, each of which can carry multiple types of traffic, to provide flexibility and scalability in the deployment of new network services. In this context, this paper presents the description of a network simulator, that addresses the key related 5G features, i.e., modeling of virtual network functions, network slicing, ability to change some network parameters during run-time, and a more realistic network traffic generation. During the demo, we deploy several slices with different types of traffic to demonstrate that our simulator can support applications like network management and orchestration. The evaluation results show that our simulator is a powerful tool for testing 5G networks.
{"title":"Demo: A Network Simulator for 5G Virtualized Networks","authors":"Oscar Delgado, B. Jaumard, Zhiyi Ding, Fadi Bishay, Vincent Bissonnette","doi":"10.1109/NetSoft54395.2022.9844042","DOIUrl":"https://doi.org/10.1109/NetSoft54395.2022.9844042","url":null,"abstract":"Current 5G and beyond research explores how to leverage network virtualization, which allows network operators to partition the network into multiple independent slices, each of which can carry multiple types of traffic, to provide flexibility and scalability in the deployment of new network services. In this context, this paper presents the description of a network simulator, that addresses the key related 5G features, i.e., modeling of virtual network functions, network slicing, ability to change some network parameters during run-time, and a more realistic network traffic generation. During the demo, we deploy several slices with different types of traffic to demonstrate that our simulator can support applications like network management and orchestration. The evaluation results show that our simulator is a powerful tool for testing 5G networks.","PeriodicalId":125799,"journal":{"name":"2022 IEEE 8th International Conference on Network Softwarization (NetSoft)","volume":"51 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131536419","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 : 2022-06-27DOI: 10.1109/NetSoft54395.2022.9844056
Menuka Perera Jayasuriya Kuranage, L. Nuaymi, A. Bouabdallah, Thomas Ferrandiz, P. Bertin
5G networks are moving towards cloudification which gives the telecom operators the flexibility to manage their networks efficiently and cost-effectively. Scaling network functions on demand is one of the advantages of using container-based deployment in cloud environments. With the continuously changing network traffic patterns due to the emerging new 5G use cases, there is a need for novel automated network resources management approach in cloud-native environments. Considering the scale and the complexity of the 5G network, managing resources is a challenge. To address this, we propose a deep learning-based resource usage forecasting approach that provides useful insights for decision-making in containerized Network Function (CNF) scaling for the Kubernetes environment. Kubernetes is a container orchestration tool that becoming popular among Telecom operators due to its simplicity. We implemented a testbed in the Kubernetes environment to generate a dataset closer to real-world data for deep learning model training and evaluated the best-performing model for resource usage forecasting. We benchmarked our approach against another deep learning-based resource usage forecasting approach which proved our method can provide a highly accurate forecast for further horizons.
{"title":"Deep learning based resource forecasting for 5G core network scaling in Kubernetes environment","authors":"Menuka Perera Jayasuriya Kuranage, L. Nuaymi, A. Bouabdallah, Thomas Ferrandiz, P. Bertin","doi":"10.1109/NetSoft54395.2022.9844056","DOIUrl":"https://doi.org/10.1109/NetSoft54395.2022.9844056","url":null,"abstract":"5G networks are moving towards cloudification which gives the telecom operators the flexibility to manage their networks efficiently and cost-effectively. Scaling network functions on demand is one of the advantages of using container-based deployment in cloud environments. With the continuously changing network traffic patterns due to the emerging new 5G use cases, there is a need for novel automated network resources management approach in cloud-native environments. Considering the scale and the complexity of the 5G network, managing resources is a challenge. To address this, we propose a deep learning-based resource usage forecasting approach that provides useful insights for decision-making in containerized Network Function (CNF) scaling for the Kubernetes environment. Kubernetes is a container orchestration tool that becoming popular among Telecom operators due to its simplicity. We implemented a testbed in the Kubernetes environment to generate a dataset closer to real-world data for deep learning model training and evaluated the best-performing model for resource usage forecasting. We benchmarked our approach against another deep learning-based resource usage forecasting approach which proved our method can provide a highly accurate forecast for further horizons.","PeriodicalId":125799,"journal":{"name":"2022 IEEE 8th International Conference on Network Softwarization (NetSoft)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129768758","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 : 2022-06-27DOI: 10.1109/NetSoft54395.2022.9844026
Ziteng Zeng, Leslie Monis, Shixiong Qi, K. Ramakrishnan
Softwarized network resident functions have been extensively used to replace purpose-built appliances. However, there is a lack of alternatives for richer network resident functionality with a seamless combination of L2/L3 Network Function Virtualization (NFV) and L4/L7 middleboxes.We propose MiddleNet, a unified L2/L3 NFV and L4/L7 middlebox framework. MiddleNet uses DPDK in L2/L3 NFV to achieve high-performance, zero-copy packet delivery. MiddleNet exploits the event-driven capabilities of extended Berkeley Packet Filter (eBPF) to build up lightweight L4/L7 middleboxes with load-proportional overheads. MiddleNet constructs complex L2/L3 NF and L4/L7 middlebox function chains with low overhead using shared memory communication. With the integration of Single Root I/O Virtualization (SR-IOV), MiddleNet supports dynamically selecting packet processing layers (L2 to L7) based on the flow. In this demo, we show MiddleNet’s operation.
{"title":"DEMO: MiddleNet: A High-Performance, Lightweight, Unified NFV & Middlebox Framework","authors":"Ziteng Zeng, Leslie Monis, Shixiong Qi, K. Ramakrishnan","doi":"10.1109/NetSoft54395.2022.9844026","DOIUrl":"https://doi.org/10.1109/NetSoft54395.2022.9844026","url":null,"abstract":"Softwarized network resident functions have been extensively used to replace purpose-built appliances. However, there is a lack of alternatives for richer network resident functionality with a seamless combination of L2/L3 Network Function Virtualization (NFV) and L4/L7 middleboxes.We propose MiddleNet, a unified L2/L3 NFV and L4/L7 middlebox framework. MiddleNet uses DPDK in L2/L3 NFV to achieve high-performance, zero-copy packet delivery. MiddleNet exploits the event-driven capabilities of extended Berkeley Packet Filter (eBPF) to build up lightweight L4/L7 middleboxes with load-proportional overheads. MiddleNet constructs complex L2/L3 NF and L4/L7 middlebox function chains with low overhead using shared memory communication. With the integration of Single Root I/O Virtualization (SR-IOV), MiddleNet supports dynamically selecting packet processing layers (L2 to L7) based on the flow. In this demo, we show MiddleNet’s operation.","PeriodicalId":125799,"journal":{"name":"2022 IEEE 8th International Conference on Network Softwarization (NetSoft)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132675476","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 : 2022-06-27DOI: 10.1109/NetSoft54395.2022.9844090
Ralf Kundel, Leonard Anderweit, Jonas Markussen, C. Griwodz, Osama Abboud, Benjamin Becker, Tobias Meuser
Hardware acceleration of network functions is essential to meet the challenging Quality of Service requirements in nowadays computer networks. Graphical Processing Units (GPU) are a widely deployed technology that can also be used for computing tasks, including acceleration of network functions. In this work, we demonstrate how commodity GPUs, which do not provide any network interfaces, can be used to accelerate network functions. Our approach leverages PCIe peer-to-peer capabilities and allows the GPU to control the network interface card directly, without any assistance from the operating system or control application. The presented evaluation results demonstrate the feasibility of our approach and its performance of up to 10 Gbit/s, even for small packets.
{"title":"Host Bypassing: Let your GPU speak Ethernet","authors":"Ralf Kundel, Leonard Anderweit, Jonas Markussen, C. Griwodz, Osama Abboud, Benjamin Becker, Tobias Meuser","doi":"10.1109/NetSoft54395.2022.9844090","DOIUrl":"https://doi.org/10.1109/NetSoft54395.2022.9844090","url":null,"abstract":"Hardware acceleration of network functions is essential to meet the challenging Quality of Service requirements in nowadays computer networks. Graphical Processing Units (GPU) are a widely deployed technology that can also be used for computing tasks, including acceleration of network functions. In this work, we demonstrate how commodity GPUs, which do not provide any network interfaces, can be used to accelerate network functions. Our approach leverages PCIe peer-to-peer capabilities and allows the GPU to control the network interface card directly, without any assistance from the operating system or control application. The presented evaluation results demonstrate the feasibility of our approach and its performance of up to 10 Gbit/s, even for small packets.","PeriodicalId":125799,"journal":{"name":"2022 IEEE 8th International Conference on Network Softwarization (NetSoft)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131965816","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：