Pub Date : 2013-11-01DOI: 10.1109/CloudNet.2013.6710556
Yuki Tanisawa, M. Yamamoto
Data center network is composed of high-speed Ethernet extended in a limited area of a data center building, so its RTT is extremely small of μsec order. In order to regulate data center network delay large part of which is queuing delay, QCN is proposed for layer 2 congestion control in IEEE 802.1Qau. QCN controls transmission rate of the sender by congestion feedback from a congested switch. QCN adopts probabilistic feedback transmission to reduce the control overhead. When the number of flows through a bottleneck link increases, some flows might receive no feedback even in congestion phase due to probabilistic feedback transmission. In this situation, queue length might be significantly fluctuated. In this paper, we propose a new delay-based congestion detection and control method. Our proposed delay-based congestion control is cooperated with the conventional QCN so as to detect and react congestion not detected by QCN.
{"title":"QCN with delay-based congestion detection for limited queue fluctuation in data center networks","authors":"Yuki Tanisawa, M. Yamamoto","doi":"10.1109/CloudNet.2013.6710556","DOIUrl":"https://doi.org/10.1109/CloudNet.2013.6710556","url":null,"abstract":"Data center network is composed of high-speed Ethernet extended in a limited area of a data center building, so its RTT is extremely small of μsec order. In order to regulate data center network delay large part of which is queuing delay, QCN is proposed for layer 2 congestion control in IEEE 802.1Qau. QCN controls transmission rate of the sender by congestion feedback from a congested switch. QCN adopts probabilistic feedback transmission to reduce the control overhead. When the number of flows through a bottleneck link increases, some flows might receive no feedback even in congestion phase due to probabilistic feedback transmission. In this situation, queue length might be significantly fluctuated. In this paper, we propose a new delay-based congestion detection and control method. Our proposed delay-based congestion control is cooperated with the conventional QCN so as to detect and react congestion not detected by QCN.","PeriodicalId":262262,"journal":{"name":"2013 IEEE 2nd International Conference on Cloud Networking (CloudNet)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130207609","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 : 2013-11-01DOI: 10.1109/CloudNet.2013.6710570
Marouen Mechtri, D. Zeghlache, E. Zekri, I. Marshall
Cloud network models have limitations in handling networking between distributed cloud resources and in providing customers the ability to control and configure networks. This paper presents a Cloud Networking Gateway (CNG) Manager for dynamic establishment of intra and inter cloud connectivity. The CNG Manager interconnects virtual machines acquired from distributed heterogeneous resources and services from multiple providers using a generic gateway. The cloud networking gateways are managed by the CNG Manager that handles allocation and configuration of the gateways according to connectivity requirements. This cloud networking solution is integrated with an operational “Cloud Broker” architecture involving multiple users and providers in order to test and validate our solution in a realistic framework.
{"title":"Inter and intra Cloud Networking Gateway as a service","authors":"Marouen Mechtri, D. Zeghlache, E. Zekri, I. Marshall","doi":"10.1109/CloudNet.2013.6710570","DOIUrl":"https://doi.org/10.1109/CloudNet.2013.6710570","url":null,"abstract":"Cloud network models have limitations in handling networking between distributed cloud resources and in providing customers the ability to control and configure networks. This paper presents a Cloud Networking Gateway (CNG) Manager for dynamic establishment of intra and inter cloud connectivity. The CNG Manager interconnects virtual machines acquired from distributed heterogeneous resources and services from multiple providers using a generic gateway. The cloud networking gateways are managed by the CNG Manager that handles allocation and configuration of the gateways according to connectivity requirements. This cloud networking solution is integrated with an operational “Cloud Broker” architecture involving multiple users and providers in order to test and validate our solution in a realistic framework.","PeriodicalId":262262,"journal":{"name":"2013 IEEE 2nd International Conference on Cloud Networking (CloudNet)","volume":"428 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132550902","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 : 2013-11-01DOI: 10.1109/CloudNet.2013.6710563
Kangkang Li, Jie Wu, Adam Blaisse
With the increasing popularity of cloud computing, the cloud datacenter suffers from both limited resources and the variation of users' requests. One important feature of cloud computing is on-demand scaling, enabling the fluctuation of one user's resource demand. However, amongst previous work concerning the virtual machine (VM) placement in datacenters, satisfying the VMs' requested resources of users is the primary objective, neglecting future demand variation. In this paper, we propose the concept of elasticity, referring to how well the datacenter can satisfy the growth of the input VMs resource demands under both the limitations of physical machines (PMs) capacities and links capacities. To consider both dimensions of the machine and bandwidth resources simultaneously, we propose our hierarchical VM placement algorithm. We also prove the optimality of our algorithm in a frequently used semi-homogeneous datacenter configuration. Furthermore, we study the heterogeneous datacenter configuration, favoring the characteristics of multi-tenant datacenters. Evaluation results validate the efficiency of our algorithm.
{"title":"Elasticity-aware virtual machine placement for cloud datacenters","authors":"Kangkang Li, Jie Wu, Adam Blaisse","doi":"10.1109/CloudNet.2013.6710563","DOIUrl":"https://doi.org/10.1109/CloudNet.2013.6710563","url":null,"abstract":"With the increasing popularity of cloud computing, the cloud datacenter suffers from both limited resources and the variation of users' requests. One important feature of cloud computing is on-demand scaling, enabling the fluctuation of one user's resource demand. However, amongst previous work concerning the virtual machine (VM) placement in datacenters, satisfying the VMs' requested resources of users is the primary objective, neglecting future demand variation. In this paper, we propose the concept of elasticity, referring to how well the datacenter can satisfy the growth of the input VMs resource demands under both the limitations of physical machines (PMs) capacities and links capacities. To consider both dimensions of the machine and bandwidth resources simultaneously, we propose our hierarchical VM placement algorithm. We also prove the optimality of our algorithm in a frequently used semi-homogeneous datacenter configuration. Furthermore, we study the heterogeneous datacenter configuration, favoring the characteristics of multi-tenant datacenters. Evaluation results validate the efficiency of our algorithm.","PeriodicalId":262262,"journal":{"name":"2013 IEEE 2nd International Conference on Cloud Networking (CloudNet)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125497656","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 : 2013-11-01DOI: 10.1109/CloudNet.2013.6710552
Ashiq Khan, W. Kiess, D. Pérez-Caparrós, J. Triay
Network virtualization presents a new networking paradigm that allows to accommodate multiple isolated virtual networks in one physical network. OpenFlow-based switches are a promising technology to achieve isolation in the transport network. However, these lack carrier-grade QoS support. On the contrary, MPLS is a scalable transport network technology with carrier-grade QoS support. It can isolate multiple QoS classes. Still, when accommodating multiple virtual networks with carrier-grade QoS support, we need to now isolate virtual networks in the same QoS class. MPLS lacks programmability to do that. In this paper, we propose a mechanism to isolate virtual networks in the same QoS class, by using OpenFlow in combination with MPLS in order to utilize the merits offered by both technologies. The core of our proposal does not need modification in the architecture from both. In addition, we also present extensions for faster response in the transport network. The evaluation of our proposal on a prototype proves the validity of our scheme as well as its usefulness in future networks.
{"title":"Quality-of-service (QoS) for virtual networks in OpenFlow MPLS transport networks","authors":"Ashiq Khan, W. Kiess, D. Pérez-Caparrós, J. Triay","doi":"10.1109/CloudNet.2013.6710552","DOIUrl":"https://doi.org/10.1109/CloudNet.2013.6710552","url":null,"abstract":"Network virtualization presents a new networking paradigm that allows to accommodate multiple isolated virtual networks in one physical network. OpenFlow-based switches are a promising technology to achieve isolation in the transport network. However, these lack carrier-grade QoS support. On the contrary, MPLS is a scalable transport network technology with carrier-grade QoS support. It can isolate multiple QoS classes. Still, when accommodating multiple virtual networks with carrier-grade QoS support, we need to now isolate virtual networks in the same QoS class. MPLS lacks programmability to do that. In this paper, we propose a mechanism to isolate virtual networks in the same QoS class, by using OpenFlow in combination with MPLS in order to utilize the merits offered by both technologies. The core of our proposal does not need modification in the architecture from both. In addition, we also present extensions for faster response in the transport network. The evaluation of our proposal on a prototype proves the validity of our scheme as well as its usefulness in future networks.","PeriodicalId":262262,"journal":{"name":"2013 IEEE 2nd International Conference on Cloud Networking (CloudNet)","volume":"268 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122753098","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 : 2013-11-01DOI: 10.1109/CloudNet.2013.6710558
Matthieu Coudron, Stefano Secci, G. Pujolle, Patrick Raad, Pascal Gallard
Cloud networking imposes new requirements in terms of connection resiliency and throughput among virtual machines, hypervisors and users. A promising direction is to exploit multipath communications, yet existing protocols have a so limited scope that performance improvements are often unreachable. Generally, multipathing adds signaling overhead and in certain conditions may in fact decrease throughput due to packet arrival disorder. At the transport layer, the most promising protocol is Multipath TCP (MPTCP), a backward compatible TCP extension allowing to balance the load on several TCP subflows, ideally following different physical paths, to maximize connection throughput. Current implementations create a full mesh between hosts IPs, which can be suboptimal. For situation when at least one end-point network is multihomed, we propose to enhance its subflow creation mechanism so that MPTCP creates an adequate number of subflows considering the underlying path diversity offered by an IP-in-IP mapping protocol, the Location/Identifier Separation Protocol (LISP). We defined and implemented a cross-layer cooperation module between MPTCP and LISP, leading to an improved version of MPTCP we name Augmented MPTCP (A-MPTCP). We evaluated A-MPTCP for a realistic Cloud access use-case scenario involving one multi-homed data-center. Results from a large-scale test bed show us that A-MPTCP can halve the transfer times with the simple addition of one additional LIS-Penabled MPTCP subflow, hence showing promising performance for Cloud communications between multi-homed users and multihomed data-centers.
云网络在虚拟机、管理程序和用户之间的连接弹性和吞吐量方面提出了新的要求。一个很有前途的方向是利用多路径通信，但是现有协议的范围非常有限，因此通常无法实现性能改进。通常，多路径增加了信令开销，并且在某些情况下实际上可能由于数据包到达混乱而降低吞吐量。在传输层，最有前途的协议是多路径TCP (MPTCP)，这是一种向后兼容的TCP扩展，允许在几个TCP子流上平衡负载，理想情况下遵循不同的物理路径，以最大限度地提高连接吞吐量。当前的实现在主机ip之间创建了一个完整的网格，这可能不是最优的。对于至少一个端点网络是多宿主的情况，我们建议增强其子流创建机制，以便MPTCP创建足够数量的子流，考虑到IP-in-IP映射协议，位置/标识符分离协议(LISP)提供的底层路径多样性。我们定义并实现了MPTCP和LISP之间的跨层合作模块，从而形成了MPTCP的改进版本，我们称之为增强型MPTCP (a -MPTCP)。我们针对一个实际的云访问用例场景评估了a- mptcp，该场景涉及一个多主数据中心。大规模测试平台的结果表明，通过简单地添加一个额外的支持lis的MPTCP子流，a- MPTCP可以将传输时间减半，因此在多宿主用户和多宿主数据中心之间的云通信中显示出有希望的性能。
{"title":"Cross-layer cooperation to boost multipath TCP performance in cloud networks","authors":"Matthieu Coudron, Stefano Secci, G. Pujolle, Patrick Raad, Pascal Gallard","doi":"10.1109/CloudNet.2013.6710558","DOIUrl":"https://doi.org/10.1109/CloudNet.2013.6710558","url":null,"abstract":"Cloud networking imposes new requirements in terms of connection resiliency and throughput among virtual machines, hypervisors and users. A promising direction is to exploit multipath communications, yet existing protocols have a so limited scope that performance improvements are often unreachable. Generally, multipathing adds signaling overhead and in certain conditions may in fact decrease throughput due to packet arrival disorder. At the transport layer, the most promising protocol is Multipath TCP (MPTCP), a backward compatible TCP extension allowing to balance the load on several TCP subflows, ideally following different physical paths, to maximize connection throughput. Current implementations create a full mesh between hosts IPs, which can be suboptimal. For situation when at least one end-point network is multihomed, we propose to enhance its subflow creation mechanism so that MPTCP creates an adequate number of subflows considering the underlying path diversity offered by an IP-in-IP mapping protocol, the Location/Identifier Separation Protocol (LISP). We defined and implemented a cross-layer cooperation module between MPTCP and LISP, leading to an improved version of MPTCP we name Augmented MPTCP (A-MPTCP). We evaluated A-MPTCP for a realistic Cloud access use-case scenario involving one multi-homed data-center. Results from a large-scale test bed show us that A-MPTCP can halve the transfer times with the simple addition of one additional LIS-Penabled MPTCP subflow, hence showing promising performance for Cloud communications between multi-homed users and multihomed data-centers.","PeriodicalId":262262,"journal":{"name":"2013 IEEE 2nd International Conference on Cloud Networking (CloudNet)","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130790982","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 : 2013-11-01DOI: 10.1109/CloudNet.2013.6710561
Kangkang Li, Huanyang Zheng, Jie Wu
Cloud computing is an emerging technology that greatly shapes our lives, where users run jobs on virtual machines (VMs) on physical machines (PMs) provided by a cloud provider, saving the investment in upfront infrastructures. Due to the heterogeneity of various jobs, different VMs on the same PM can have different job completion times. Meanwhile, the PMs are also heterogeneous. Therefore, different VM placements have different job completion times. Our objective is to minimize the total job completion time of the input VM requests through a reasonable VM placement schedule. This problem is NP-hard, since it can be reduced to a knapsack problem. We propose an off-line VM placement method through emulated VM migration, while the on-line VM placement is solved by a real VM migration process. The migration algorithm is a heuristic approach, where we place the VM to its best PM directly, as long as it has enough capacity. Otherwise, if the migration constraint is satisfied, we migrate another VM from this PM to accommodate the new VM. Furthermore, we study a hybrid scheme where a batch is employed to accept upcoming VMs for the on-line scenario. Evaluation results prove the high efficiency of our algorithms.
{"title":"Migration-based virtual machine placement in cloud systems","authors":"Kangkang Li, Huanyang Zheng, Jie Wu","doi":"10.1109/CloudNet.2013.6710561","DOIUrl":"https://doi.org/10.1109/CloudNet.2013.6710561","url":null,"abstract":"Cloud computing is an emerging technology that greatly shapes our lives, where users run jobs on virtual machines (VMs) on physical machines (PMs) provided by a cloud provider, saving the investment in upfront infrastructures. Due to the heterogeneity of various jobs, different VMs on the same PM can have different job completion times. Meanwhile, the PMs are also heterogeneous. Therefore, different VM placements have different job completion times. Our objective is to minimize the total job completion time of the input VM requests through a reasonable VM placement schedule. This problem is NP-hard, since it can be reduced to a knapsack problem. We propose an off-line VM placement method through emulated VM migration, while the on-line VM placement is solved by a real VM migration process. The migration algorithm is a heuristic approach, where we place the VM to its best PM directly, as long as it has enough capacity. Otherwise, if the migration constraint is satisfied, we migrate another VM from this PM to accommodate the new VM. Furthermore, we study a hybrid scheme where a batch is employed to accept upcoming VMs for the on-line scenario. Evaluation results prove the high efficiency of our algorithms.","PeriodicalId":262262,"journal":{"name":"2013 IEEE 2nd International Conference on Cloud Networking (CloudNet)","volume":"254 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122934055","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 : 2013-11-01DOI: 10.1109/CloudNet.2013.6710557
S. Ha, Dino Lopez Pacheco, G. Urvoy-Keller
Virtualization is heavily used in modern private and public data centers. By enabling the consolidation of many virtual machines (VMs) in a physical server, virtualization blurs the frontier between networking and system. We have investigated in a previous work the various prices due to virtualization when a packet is moved from one VM to the physical NIC, by focusing on a pure UDP scenario and multiple competing VMs. We estimated how the inter-packet delay moves away from the optimal point as the number of VMs increases. In the present article we report the impact of virtualization over a whole data stream controlled by the TCP protocol. Our findings reveal that the impact of the additional inter-packet delay is negligible on the performance of TCP. Indeed, virtualized systems have similar goodput levels to the ones of natives systems. Moreover, a less expected result show that virtualized systems improve the fairness between flows, even when such flows are streamed from a same VM. The bad consequences of the virtualization concern mainly bugs in the implementation of the networking function, which can potentially introduce spurious retransmissions and information leaks.
{"title":"Networking in a virtualized environment: The TCP case","authors":"S. Ha, Dino Lopez Pacheco, G. Urvoy-Keller","doi":"10.1109/CloudNet.2013.6710557","DOIUrl":"https://doi.org/10.1109/CloudNet.2013.6710557","url":null,"abstract":"Virtualization is heavily used in modern private and public data centers. By enabling the consolidation of many virtual machines (VMs) in a physical server, virtualization blurs the frontier between networking and system. We have investigated in a previous work the various prices due to virtualization when a packet is moved from one VM to the physical NIC, by focusing on a pure UDP scenario and multiple competing VMs. We estimated how the inter-packet delay moves away from the optimal point as the number of VMs increases. In the present article we report the impact of virtualization over a whole data stream controlled by the TCP protocol. Our findings reveal that the impact of the additional inter-packet delay is negligible on the performance of TCP. Indeed, virtualized systems have similar goodput levels to the ones of natives systems. Moreover, a less expected result show that virtualized systems improve the fairness between flows, even when such flows are streamed from a same VM. The bad consequences of the virtualization concern mainly bugs in the implementation of the networking function, which can potentially introduce spurious retransmissions and information leaks.","PeriodicalId":262262,"journal":{"name":"2013 IEEE 2nd International Conference on Cloud Networking (CloudNet)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122756103","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 : 2013-11-01DOI: 10.1109/CloudNet.2013.6710584
Haiyang Qian, Xin Huang, Clark Chen
Networking Virtualization abstracts the underlying networking away from the tenants in multiple-tenant cloud data center (DC). Thus the tenants can apply their policies, manage their address space, group and migrate their VMs independently. The DCs are interconnected by Wide Area Network (WAN). To provide virtual network across multiple DCs is a desirable feature. The bandwidth of WAN is, however, a scarce resource. And the different encapsulation of packet in the DC and WAN causes the mechanism to handle Quality of Service (QoS) are different. In this paper, we leverage Software-Define Networking to achieve network virtualization in the DC orchestrated by OpenStack and convey the application level QoS from the DC to the WAN. We design an end-to-end DC and WAN orchestration architecture to extend the network virtualization across multiple DCs with application level QoS awareness and consistency.
{"title":"SWAN: End-to-end orchestration for cloud network and WAN","authors":"Haiyang Qian, Xin Huang, Clark Chen","doi":"10.1109/CloudNet.2013.6710584","DOIUrl":"https://doi.org/10.1109/CloudNet.2013.6710584","url":null,"abstract":"Networking Virtualization abstracts the underlying networking away from the tenants in multiple-tenant cloud data center (DC). Thus the tenants can apply their policies, manage their address space, group and migrate their VMs independently. The DCs are interconnected by Wide Area Network (WAN). To provide virtual network across multiple DCs is a desirable feature. The bandwidth of WAN is, however, a scarce resource. And the different encapsulation of packet in the DC and WAN causes the mechanism to handle Quality of Service (QoS) are different. In this paper, we leverage Software-Define Networking to achieve network virtualization in the DC orchestrated by OpenStack and convey the application level QoS from the DC to the WAN. We design an end-to-end DC and WAN orchestration architecture to extend the network virtualization across multiple DCs with application level QoS awareness and consistency.","PeriodicalId":262262,"journal":{"name":"2013 IEEE 2nd International Conference on Cloud Networking (CloudNet)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126575248","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 : 2013-11-01DOI: 10.1109/CloudNet.2013.6710568
Ondrej Tomanek, L. Kencl
Latency is an important, yet often underestimated aspect of the nascent Cloud-Computing scenario. A cloud service based on processing in remote datacenters may exhibit latency and jitter which may be a compound result of many various components of the remote computation and intermediate communication. Our broad vision is to design and develop tools to monitor, model and optimize the global cloud-service latency. To this end, we introduce CLAudit, a prototype planetary-scale cloud-latency auditing platform. It utilizes the experimental PlanetLab network to place globally distributed probes that periodically measure cloud-service latency at various layers of the communication stack. We present CLAudit architecture in detail and show initial test-measurements of the MicrosoftWindows Azure cloud service, demonstrating the platform's practical usefulness by showcasing a few discovered anomalous results in the cloud-service latency measurements.
{"title":"CLAudit: Planetary-scale cloud latency auditing platform","authors":"Ondrej Tomanek, L. Kencl","doi":"10.1109/CloudNet.2013.6710568","DOIUrl":"https://doi.org/10.1109/CloudNet.2013.6710568","url":null,"abstract":"Latency is an important, yet often underestimated aspect of the nascent Cloud-Computing scenario. A cloud service based on processing in remote datacenters may exhibit latency and jitter which may be a compound result of many various components of the remote computation and intermediate communication. Our broad vision is to design and develop tools to monitor, model and optimize the global cloud-service latency. To this end, we introduce CLAudit, a prototype planetary-scale cloud-latency auditing platform. It utilizes the experimental PlanetLab network to place globally distributed probes that periodically measure cloud-service latency at various layers of the communication stack. We present CLAudit architecture in detail and show initial test-measurements of the MicrosoftWindows Azure cloud service, demonstrating the platform's practical usefulness by showcasing a few discovered anomalous results in the cloud-service latency measurements.","PeriodicalId":262262,"journal":{"name":"2013 IEEE 2nd International Conference on Cloud Networking (CloudNet)","volume":"44 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123527431","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 : 2013-11-01DOI: 10.1109/CloudNet.2013.6710576
Sonia Belhareth, L. Sassatelli, D. Collange, Dino Lopez Pacheco, G. Urvoy-Keller
Cloud networking typically leads to scenarii where a large number of TCP connections share a common bottleneck link. In this paper, we focus on the case of TCP Cubic, which is the default TCP version in the Linux kernel. TCP Cubic is designed to better utilize high bandwidth-delay product path in an IP network. To do so, Cubic modifies the linear window growth function of legacy TCP standards, e.g., New Reno, to be a cubic function. Our objective in this work is to assess the performance of TCP Cubic in a cloud setting with a large number of long-lived TCP flows.We rely on a mean-field approach leading to a fluid model to analyze the performance of Cubic. After a careful validation of the model through comparisons with ns-2, we evaluate the efficiency and fairness of Cubic as compared to that of New Reno for a set of typical cloud networking scenarii.
{"title":"Understanding TCP cubic performance in the cloud: A mean-field approach","authors":"Sonia Belhareth, L. Sassatelli, D. Collange, Dino Lopez Pacheco, G. Urvoy-Keller","doi":"10.1109/CloudNet.2013.6710576","DOIUrl":"https://doi.org/10.1109/CloudNet.2013.6710576","url":null,"abstract":"Cloud networking typically leads to scenarii where a large number of TCP connections share a common bottleneck link. In this paper, we focus on the case of TCP Cubic, which is the default TCP version in the Linux kernel. TCP Cubic is designed to better utilize high bandwidth-delay product path in an IP network. To do so, Cubic modifies the linear window growth function of legacy TCP standards, e.g., New Reno, to be a cubic function. Our objective in this work is to assess the performance of TCP Cubic in a cloud setting with a large number of long-lived TCP flows.We rely on a mean-field approach leading to a fluid model to analyze the performance of Cubic. After a careful validation of the model through comparisons with ns-2, we evaluate the efficiency and fairness of Cubic as compared to that of New Reno for a set of typical cloud networking scenarii.","PeriodicalId":262262,"journal":{"name":"2013 IEEE 2nd International Conference on Cloud Networking (CloudNet)","volume":"53 10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132484469","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：