Pub Date : 2022-10-31DOI: 10.23919/CNSM55787.2022.9964896
Edenilson Jônatas dos Passos, Adriano Fiorese
Cloud computing and video streaming services have been in constant expansion in recent years. Along with it, the demand for computing resources has also increased significantly. In this context, monitoring the use of these resources is crucial to maintain a satisfactory level of Quality of Service and, consequently, Quality of Experience, especially in video transmission services. This work discusses a new method of monitoring resources and quality of service metrics on content servers involving CPU utilization and server throughput, which is obtained in a distributed way. For that, a distributed collector system that is based on a modified version of the ring election algorithm is developed to retrieve the Quality of Service metrics in each server. Evaluation experiment results show that there are no performance gains on the system such as the content loading faster for the user, there are however, improvements in terms of the whole system scalability. The greater the number of servers for monitoring, the better the approach is compared to the traditional method of monitoring resources through request and response.
{"title":"Monitoring Metrics for Load Balancing over Video Content Distribution Servers","authors":"Edenilson Jônatas dos Passos, Adriano Fiorese","doi":"10.23919/CNSM55787.2022.9964896","DOIUrl":"https://doi.org/10.23919/CNSM55787.2022.9964896","url":null,"abstract":"Cloud computing and video streaming services have been in constant expansion in recent years. Along with it, the demand for computing resources has also increased significantly. In this context, monitoring the use of these resources is crucial to maintain a satisfactory level of Quality of Service and, consequently, Quality of Experience, especially in video transmission services. This work discusses a new method of monitoring resources and quality of service metrics on content servers involving CPU utilization and server throughput, which is obtained in a distributed way. For that, a distributed collector system that is based on a modified version of the ring election algorithm is developed to retrieve the Quality of Service metrics in each server. Evaluation experiment results show that there are no performance gains on the system such as the content loading faster for the user, there are however, improvements in terms of the whole system scalability. The greater the number of servers for monitoring, the better the approach is compared to the traditional method of monitoring resources through request and response.","PeriodicalId":232521,"journal":{"name":"2022 18th International Conference on Network and Service Management (CNSM)","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133652099","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 : 2022-10-31DOI: 10.23919/CNSM55787.2022.9964722
Asterios Mpatziakas, Anastasios Sinanis, Iosif Hamlatzis, A. Drosou, D. Tzovaras
5G architectures will utilize the virtualization of the network functions (VNF) and the use of Multi-access edge computing (MEC) to gain multiple benefits such as simpler service orchestration, while simultaneously covering diverse use cases even with strict performance requirements. 5G service orchestration mechanisms will need to allow more efficient and flexible network deployment and operations in a resource-efficient and delay-sensitive manner. A field that is expected to be greatly boosted by these advances, is Cellular Vehicle to Everything communications. 5G will enable cooperative, connected and automated mobility services, which are often are safety critical while also having stringent delay requirements. This paper, proposes a mechanism that predicts the future position of a vehicle moving in both urban and/or highway environments. Based on this knowledge, it decides on the optimal position of VNFs so that the allocation of network resources can be preemptively requested. The objective of this mechanism is to ensure the uninterrupted, continuous connections of the vehicles, resulting in minimal or no service interruption time while ensuring an optimal utilization of Edge Cloud and MEC resources.
{"title":"AI-Based mechanism for the Predictive Resource Allocation of V2X related Network Services","authors":"Asterios Mpatziakas, Anastasios Sinanis, Iosif Hamlatzis, A. Drosou, D. Tzovaras","doi":"10.23919/CNSM55787.2022.9964722","DOIUrl":"https://doi.org/10.23919/CNSM55787.2022.9964722","url":null,"abstract":"5G architectures will utilize the virtualization of the network functions (VNF) and the use of Multi-access edge computing (MEC) to gain multiple benefits such as simpler service orchestration, while simultaneously covering diverse use cases even with strict performance requirements. 5G service orchestration mechanisms will need to allow more efficient and flexible network deployment and operations in a resource-efficient and delay-sensitive manner. A field that is expected to be greatly boosted by these advances, is Cellular Vehicle to Everything communications. 5G will enable cooperative, connected and automated mobility services, which are often are safety critical while also having stringent delay requirements. This paper, proposes a mechanism that predicts the future position of a vehicle moving in both urban and/or highway environments. Based on this knowledge, it decides on the optimal position of VNFs so that the allocation of network resources can be preemptively requested. The objective of this mechanism is to ensure the uninterrupted, continuous connections of the vehicles, resulting in minimal or no service interruption time while ensuring an optimal utilization of Edge Cloud and MEC resources.","PeriodicalId":232521,"journal":{"name":"2022 18th International Conference on Network and Service Management (CNSM)","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123497536","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 : 2022-10-31DOI: 10.23919/CNSM55787.2022.9964740
Imed Ghnaya, T. Ahmed, M. Mosbah, H. Aniss
Cooperative Perception (CP) allows Connected and Autonomous Vehicles (CAVs) to enhance their Environmental Awareness (EA) by sharing locally perceived objects through CP messages (CPMs). European Telecommunications Standards Institute (ETSI) has recently defined a set of CPM generation rules to achieve a trade-off between EA and Channel Busy Ratio (CBR) despite massive perception data. Nonetheless, these rules still lack the consideration of information usefulness, resulting in a considerable volume of useless information transmitted in the CP network. This limitation could increase CBR and thus decrease EA due to the loss of CPMs in the network. This paper introduces CloudAC-IU, a cloud-based deep reinforcement learning approach to lean CAVs to maximize perception information usefulness in the network. Simulation results highlight that the CloudAC-IU enhances EA by decreasing CBR and increasing CPM reception for CAVs compared to state-of-the-art works.
{"title":"Maximizing Information Usefulness in Vehicular CP Networks Using Actor-Critic Reinforcement Learning","authors":"Imed Ghnaya, T. Ahmed, M. Mosbah, H. Aniss","doi":"10.23919/CNSM55787.2022.9964740","DOIUrl":"https://doi.org/10.23919/CNSM55787.2022.9964740","url":null,"abstract":"Cooperative Perception (CP) allows Connected and Autonomous Vehicles (CAVs) to enhance their Environmental Awareness (EA) by sharing locally perceived objects through CP messages (CPMs). European Telecommunications Standards Institute (ETSI) has recently defined a set of CPM generation rules to achieve a trade-off between EA and Channel Busy Ratio (CBR) despite massive perception data. Nonetheless, these rules still lack the consideration of information usefulness, resulting in a considerable volume of useless information transmitted in the CP network. This limitation could increase CBR and thus decrease EA due to the loss of CPMs in the network. This paper introduces CloudAC-IU, a cloud-based deep reinforcement learning approach to lean CAVs to maximize perception information usefulness in the network. Simulation results highlight that the CloudAC-IU enhances EA by decreasing CBR and increasing CPM reception for CAVs compared to state-of-the-art works.","PeriodicalId":232521,"journal":{"name":"2022 18th International Conference on Network and Service Management (CNSM)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129980224","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 : 2022-10-31DOI: 10.23919/CNSM55787.2022.9964607
Sanaa Ghandi, Alexandre Reiffers-Masson, Sandrine Vaton, T. Chonavel
In network monitoring, delays are of great use when it comes to QoS or content distributed services. However, it is often impossible to have access to all the delay measurements within a network. This can be due to network failures or to established measurement policies. For these reasons, delay matrix completion techniques are important for an optimal network monitoring service. In this paper, we formulate the completion problem as a neural collaborative filtering problem by testing two different architectures, generalized matrix factorization and multi-layer perceptron. We evaluate these methods on two different datasets: a synthetic one generated by an autonomous system simulator, and a real-world dataset from Ripe Atlas platform. Finally, a comparative study is conducted between these neural collaborative filtering methods and standard approaches.
{"title":"Neural Collaborative Filtering for Network Delay Matrix Completion","authors":"Sanaa Ghandi, Alexandre Reiffers-Masson, Sandrine Vaton, T. Chonavel","doi":"10.23919/CNSM55787.2022.9964607","DOIUrl":"https://doi.org/10.23919/CNSM55787.2022.9964607","url":null,"abstract":"In network monitoring, delays are of great use when it comes to QoS or content distributed services. However, it is often impossible to have access to all the delay measurements within a network. This can be due to network failures or to established measurement policies. For these reasons, delay matrix completion techniques are important for an optimal network monitoring service. In this paper, we formulate the completion problem as a neural collaborative filtering problem by testing two different architectures, generalized matrix factorization and multi-layer perceptron. We evaluate these methods on two different datasets: a synthetic one generated by an autonomous system simulator, and a real-world dataset from Ripe Atlas platform. Finally, a comparative study is conducted between these neural collaborative filtering methods and standard approaches.","PeriodicalId":232521,"journal":{"name":"2022 18th International Conference on Network and Service Management (CNSM)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132373092","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 : 2022-10-31DOI: 10.23919/CNSM55787.2022.9964617
Lukáš Melcher, Karel Hynek, T. Čejka
DNS over TLS (DoT) is one of the approaches for private DNS resolution, which has already gained support by open resolvers. Moreover, DoT is used by default in Android operating systems. This study investigates the possibility of creating DNS covert channels using DoT, which is a security threat that benefits from the increased privacy of encrypted communication. We evaluated the performance and usability of DoT tunnels created via commonly used resolvers. Our results show that the performance characteristics of DoT tunnels differ vastly depending on the used DoT resolver; however, the creation of a DoT tunnel is possible, reaching speeds up to 232 Kbps. Moreover, we successfully transferred data via DoT servers claiming Anti-Virus protection and family-friendly content.
DNS over TLS (DoT)是私有DNS解析的一种方法，已经得到了开放解析器的支持。此外，Android操作系统默认使用DoT。本研究调查了使用DoT创建DNS隐蔽通道的可能性，这是一种安全威胁，受益于加密通信的隐私性增加。我们评估了通过常用解析器创建的DoT隧道的性能和可用性。我们的研究结果表明，DoT隧道的性能特征因使用的DoT解析器而有很大差异;然而，建立一个DoT隧道是可能的，达到232 Kbps的速度。此外，我们成功地通过DoT服务器传输数据，声称防病毒保护和家庭友好的内容。
{"title":"Tunneling through DNS over TLS providers","authors":"Lukáš Melcher, Karel Hynek, T. Čejka","doi":"10.23919/CNSM55787.2022.9964617","DOIUrl":"https://doi.org/10.23919/CNSM55787.2022.9964617","url":null,"abstract":"DNS over TLS (DoT) is one of the approaches for private DNS resolution, which has already gained support by open resolvers. Moreover, DoT is used by default in Android operating systems. This study investigates the possibility of creating DNS covert channels using DoT, which is a security threat that benefits from the increased privacy of encrypted communication. We evaluated the performance and usability of DoT tunnels created via commonly used resolvers. Our results show that the performance characteristics of DoT tunnels differ vastly depending on the used DoT resolver; however, the creation of a DoT tunnel is possible, reaching speeds up to 232 Kbps. Moreover, we successfully transferred data via DoT servers claiming Anti-Virus protection and family-friendly content.","PeriodicalId":232521,"journal":{"name":"2022 18th International Conference on Network and Service Management (CNSM)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126691890","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}
Federated learning (FL) has emerged as a popular paradigm for distributed machine learning among vast clients. Unfortunately, resource-constrained clients often fail to participate in FL because they cannot pay for the memory resources required for model training due to their limited memory or bandwidth. Split federated learning (SFL) is a novel FL framework in which clients commit intermediate results of model training to a cloud server for client-server collaborative training of models, making resource-constrained clients also eligible for FL. However, existing SFL frameworks mostly require frequent communication with the cloud server to exchange intermediate results and model parameters, which results in significant communication overhead and elongated training time. In particular, this can be exacerbated by the imbalanced data distributions of clients. To tackle this issue, we propose HSFL, a hierarchical split federated learning framework that efficiently trains SFL model through hierarchical organization participants. Under the HSFL framework, we formulate a Cloud Aggregation Time Minimization (CATM) problem to minimize the global training time and design a light-weight client assignment algorithm based on dynamic programming to solve it. Moreover, we develop a self-adaption approach to cope with the dynamic computational resources of clients. Finally, we implement and evaluate HSFL on various real-world training tasks, elaborating on its effectiveness and superiority in terms of efficiency and accuracy compared to baselines.
联邦学习(FL)已经成为分布式机器学习的一个流行范例。不幸的是，资源受限的客户端经常无法参与FL，因为由于有限的内存或带宽，他们无法支付模型训练所需的内存资源。拆分联邦学习(Split federated learning, SFL)是一种新颖的模型学习框架，客户端将模型训练的中间结果提交给云服务器进行客户端-服务器协同训练模型，使得资源有限的客户端也可以进行模型学习。然而，现有的SFL框架大多需要与云服务器频繁通信以交换中间结果和模型参数，这导致通信开销大，训练时间长。特别是，客户机数据分布的不平衡可能会加剧这种情况。为了解决这个问题，我们提出了HSFL，一个分层分裂联邦学习框架，通过分层组织参与者有效地训练SFL模型。在HSFL框架下，提出了最小化全局训练时间的CATM (Cloud Aggregation Time Minimization)问题，并设计了基于动态规划的轻量级客户端分配算法来解决该问题。此外，我们还开发了一种自适应方法来处理客户端动态计算资源。最后，我们在各种现实世界的训练任务中实施和评估HSFL，阐述了与基线相比，HSFL在效率和准确性方面的有效性和优越性。
{"title":"HSFL: An Efficient Split Federated Learning Framework via Hierarchical Organization","authors":"Tengxi Xia, Yongheng Deng, Sheng Yue, Junyi He, Ju Ren, Yaoxue Zhang","doi":"10.23919/CNSM55787.2022.9964646","DOIUrl":"https://doi.org/10.23919/CNSM55787.2022.9964646","url":null,"abstract":"Federated learning (FL) has emerged as a popular paradigm for distributed machine learning among vast clients. Unfortunately, resource-constrained clients often fail to participate in FL because they cannot pay for the memory resources required for model training due to their limited memory or bandwidth. Split federated learning (SFL) is a novel FL framework in which clients commit intermediate results of model training to a cloud server for client-server collaborative training of models, making resource-constrained clients also eligible for FL. However, existing SFL frameworks mostly require frequent communication with the cloud server to exchange intermediate results and model parameters, which results in significant communication overhead and elongated training time. In particular, this can be exacerbated by the imbalanced data distributions of clients. To tackle this issue, we propose HSFL, a hierarchical split federated learning framework that efficiently trains SFL model through hierarchical organization participants. Under the HSFL framework, we formulate a Cloud Aggregation Time Minimization (CATM) problem to minimize the global training time and design a light-weight client assignment algorithm based on dynamic programming to solve it. Moreover, we develop a self-adaption approach to cope with the dynamic computational resources of clients. Finally, we implement and evaluate HSFL on various real-world training tasks, elaborating on its effectiveness and superiority in terms of efficiency and accuracy compared to baselines.","PeriodicalId":232521,"journal":{"name":"2022 18th International Conference on Network and Service Management (CNSM)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115826270","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 : 2022-10-31DOI: 10.23919/CNSM55787.2022.9964941
S. Patri, Shabnam Sultana, Michael Dürre, Saquib Amjad, Aijana Schumann, A. Autenrieth, J. Elbers, T. Bauschert, C. M. Machuca
Fully disaggregated device deployments in optical networks propose to drive down network upgrade costs. These devices are managed by open-source control plane solutions for multi-vendor interoperability, which need to be tested in a simulation environment. We demonstrate a cloud-based solution which deploys 69 OpenROADM-based containerized optical networking elements, thereby simulating a nation-wide fully disaggregated optical transport network. Further, the planning, orchestration, and restoration of optical services can be decoupled from the simulated network, by using Transport Layer Security (TLS) enabled North-Bound REST APIs exposed by OpenDayLight TransportPCE, which is an open-source optical domain controller.
{"title":"Open-Source Service Management for a Fully Disaggregated Optical Network Simulation","authors":"S. Patri, Shabnam Sultana, Michael Dürre, Saquib Amjad, Aijana Schumann, A. Autenrieth, J. Elbers, T. Bauschert, C. M. Machuca","doi":"10.23919/CNSM55787.2022.9964941","DOIUrl":"https://doi.org/10.23919/CNSM55787.2022.9964941","url":null,"abstract":"Fully disaggregated device deployments in optical networks propose to drive down network upgrade costs. These devices are managed by open-source control plane solutions for multi-vendor interoperability, which need to be tested in a simulation environment. We demonstrate a cloud-based solution which deploys 69 OpenROADM-based containerized optical networking elements, thereby simulating a nation-wide fully disaggregated optical transport network. Further, the planning, orchestration, and restoration of optical services can be decoupled from the simulated network, by using Transport Layer Security (TLS) enabled North-Bound REST APIs exposed by OpenDayLight TransportPCE, which is an open-source optical domain controller.","PeriodicalId":232521,"journal":{"name":"2022 18th International Conference on Network and Service Management (CNSM)","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125307119","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 : 2022-10-31DOI: 10.23919/CNSM55787.2022.9964491
Cristian Bermudez Serna, C. M. Machuca
Software-Defined Networking (SDN) has redefined the architectural blueprint for designing networks suitable for future applications. Today, the idea of a centralized control plane managing its underlying resources is common for architectures in mobile and industrial networks. Guaranteeing resources availability for optimal operation of the control plane is of vital importance in SDN, since compromising the controller may result in an unforeseen behaviour in the data plane. This work focuses on the SDN reactive configuration mechanism, that although originally designed for the efficient handling of changing conditions in the data plane, it can be easily misused to overload the control plane. Aiming at addressing this problem, the PDP (Programmable Data Plane)-based Controller Protection Protocol (PCPP) is presented. This protocol introduces a mechanism that efficiently filters spoofed requests at the network edge. In PCPP, end-stations require to solve a challenge before sending any connection request to the controller. The challenge answer is checked at the edge switches, which only forward valid requests to the controller. PCPP is implemented using P4, a language for programming PDP-capable devices, and its evaluation is carried out using BMv2 software switches. The results demonstrate the effectiveness of PCPP at protecting bandwidth and processing resources in the control plane against spoofed requests. A comparison against an state-of-the-art alternative not only highlights the higher efficiency of PCPP, but also its application flexibility.
{"title":"Preventing Control Plane Overload in SDN Networks with Programmable Data Planes","authors":"Cristian Bermudez Serna, C. M. Machuca","doi":"10.23919/CNSM55787.2022.9964491","DOIUrl":"https://doi.org/10.23919/CNSM55787.2022.9964491","url":null,"abstract":"Software-Defined Networking (SDN) has redefined the architectural blueprint for designing networks suitable for future applications. Today, the idea of a centralized control plane managing its underlying resources is common for architectures in mobile and industrial networks. Guaranteeing resources availability for optimal operation of the control plane is of vital importance in SDN, since compromising the controller may result in an unforeseen behaviour in the data plane. This work focuses on the SDN reactive configuration mechanism, that although originally designed for the efficient handling of changing conditions in the data plane, it can be easily misused to overload the control plane. Aiming at addressing this problem, the PDP (Programmable Data Plane)-based Controller Protection Protocol (PCPP) is presented. This protocol introduces a mechanism that efficiently filters spoofed requests at the network edge. In PCPP, end-stations require to solve a challenge before sending any connection request to the controller. The challenge answer is checked at the edge switches, which only forward valid requests to the controller. PCPP is implemented using P4, a language for programming PDP-capable devices, and its evaluation is carried out using BMv2 software switches. The results demonstrate the effectiveness of PCPP at protecting bandwidth and processing resources in the control plane against spoofed requests. A comparison against an state-of-the-art alternative not only highlights the higher efficiency of PCPP, but also its application flexibility.","PeriodicalId":232521,"journal":{"name":"2022 18th International Conference on Network and Service Management (CNSM)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126597385","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 : 2022-10-31DOI: 10.23919/CNSM55787.2022.9964863
Kaan Aykurt, Johannes Zerwas, Andreas Blenk, W. Kellerer
Today’s data centers are hosting various applications under the same roof. The diversity among deployed applications leads to a complex traffic mix in Data Center Networks (DCNs). Reconfigurable Data Center Networks (RD-CNs) have been designed to fulfill the demanding requirements of ever-changing data center traffic. However, they pose new challenges for network traffic engineering, e.g., interference between reconfigurations and congestion control (CC). This raises a fundamental research problem: can the current transport layer protocols handle frequent network updates?; This paper focuses on the Transmission Control Protocol (TCP) and presents a measurement study of TCP variants in RDCNs. The quantitative analysis of the measurements shows that migrated flows suffer from frequent reconfigurations. The effect of reconfigurations on the cost, e.g. increased Flow Completion Time (FCT), depending on the traffic mix is modeled with Machine Learning (ML) methods. The availability of such a model will provide insights into the relationship between the reconfiguration settings and the FCT. Our model explains 88% of the variance in the FCT increase under different reconfiguration settings.
{"title":"On the Performance of TCP in Reconfigurable Data Center Networks","authors":"Kaan Aykurt, Johannes Zerwas, Andreas Blenk, W. Kellerer","doi":"10.23919/CNSM55787.2022.9964863","DOIUrl":"https://doi.org/10.23919/CNSM55787.2022.9964863","url":null,"abstract":"Today’s data centers are hosting various applications under the same roof. The diversity among deployed applications leads to a complex traffic mix in Data Center Networks (DCNs). Reconfigurable Data Center Networks (RD-CNs) have been designed to fulfill the demanding requirements of ever-changing data center traffic. However, they pose new challenges for network traffic engineering, e.g., interference between reconfigurations and congestion control (CC). This raises a fundamental research problem: can the current transport layer protocols handle frequent network updates?; This paper focuses on the Transmission Control Protocol (TCP) and presents a measurement study of TCP variants in RDCNs. The quantitative analysis of the measurements shows that migrated flows suffer from frequent reconfigurations. The effect of reconfigurations on the cost, e.g. increased Flow Completion Time (FCT), depending on the traffic mix is modeled with Machine Learning (ML) methods. The availability of such a model will provide insights into the relationship between the reconfiguration settings and the FCT. Our model explains 88% of the variance in the FCT increase under different reconfiguration settings.","PeriodicalId":232521,"journal":{"name":"2022 18th International Conference on Network and Service Management (CNSM)","volume":"51 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126793329","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 : 2022-10-31DOI: 10.23919/CNSM55787.2022.9965060
Diarmuid Corcoran, P. Kreuger, Magnus Boman
As design, deployment and operation complexity increase in mobile systems, adaptive self-learning techniques have become essential enablers in mitigation and control of the complexity problem. Artificial intelligence and, in particular, reinforcement learning has shown great potential in learning complex tasks through observations. The majority of ongoing reinforcement learning research activities focus on single-agent problem settings with an assumption of accessibility to a globally observable state and action space. In many real-world settings, such as LTE or 5G, decision making is distributed and there is often only local accessibility to the state space. In such settings, multi-agent learning may be preferable, with the added challenge of ensuring that all agents collaboratively work towards achieving a common goal. We present a novel cooperative and distributed actor-critic multi-agent reinforcement learning algorithm. We claim the approach is sample efficient, both in terms of selecting observation samples and in terms of assignment of credit between subsets of collaborating agents.
{"title":"A Sample Efficient Multi-Agent Approach to Continuous Reinforcement Learning","authors":"Diarmuid Corcoran, P. Kreuger, Magnus Boman","doi":"10.23919/CNSM55787.2022.9965060","DOIUrl":"https://doi.org/10.23919/CNSM55787.2022.9965060","url":null,"abstract":"As design, deployment and operation complexity increase in mobile systems, adaptive self-learning techniques have become essential enablers in mitigation and control of the complexity problem. Artificial intelligence and, in particular, reinforcement learning has shown great potential in learning complex tasks through observations. The majority of ongoing reinforcement learning research activities focus on single-agent problem settings with an assumption of accessibility to a globally observable state and action space. In many real-world settings, such as LTE or 5G, decision making is distributed and there is often only local accessibility to the state space. In such settings, multi-agent learning may be preferable, with the added challenge of ensuring that all agents collaboratively work towards achieving a common goal. We present a novel cooperative and distributed actor-critic multi-agent reinforcement learning algorithm. We claim the approach is sample efficient, both in terms of selecting observation samples and in terms of assignment of credit between subsets of collaborating agents.","PeriodicalId":232521,"journal":{"name":"2022 18th International Conference on Network and Service Management (CNSM)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125780670","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：