光电路在分组/电路混合数据中心网络中的应用

Y. Ben-Itzhak, C. Caba, José Soler
{"title":"光电路在分组/电路混合数据中心网络中的应用","authors":"Y. Ben-Itzhak, C. Caba, José Soler","doi":"10.1145/2928275.2933284","DOIUrl":null,"url":null,"abstract":"Existing Data Center Networks (DCNs) continue to evolve to keep up with application requirements in terms of bandwidth, latency, agility, etc. According to the updated release of the Cisco Global Cloud Index [1], by 2019, more than 86% of traffic workloads will be processed by cloud DCs. Traditional DCNs, which are based on electrical packet switching (EPS) with hierarchical, tree-like topologies can no longer support future cloud traffic requirements in terms of dynamicity, bandwidth and latency. Hence, existing DCNs can be enhanced with OCS (Optical Circuit Switching), which provides high bandwidth, low latency and low power consumption [2], giving rise to hybrid OCS-EPS topologies. In this research, we assess a virtualized, hybrid, flat DCN topology consisting of a single layer of high radix ToR (Top of Rack) switches, interconnected with each other and through an OCS plane. The benefit of such flat topology is twofold: 1) In terms of bandwidth, over-subscription is reduced, and bisection bandwidth is increased; and 2) In terms of latency, the diameter (longest path) of topology is reduced. Moreover, we present new algorithms and orchestration functionality to detect and offload suitable flows (e.g. elephant flows) from the EPS to the OCS plane. Our DC architecture consists of hybrid EPS-OCS DCN, an Openflow(OF) based control plane, and an orchestration layer. Our orchestration layer decouples the elephant flows detection from the rerouting decision logic in the DCN. Specifically, the elephant flows detection is done by flow tagging in the hypervisor, while the flow rerouting is executed at the EPSs, which are connected directly to the OCS. Hence, it provides a more efficient, scalable, and easy to configure architecture as compared to existing hybrid solutions. The orchestrator monitors the ToR switches by sFlow and detects high volume traffic between two ToRs, exceeding a given bandwidth threshold. Such traffic may consist of either few elephant flows or many mice flows. To further increase the optical circuit utilization, we introduce two types of optical circuits: 1) private circuit, presented in existing solutions, is utilized only by flows that originate and end at the ToR switches connected to the circuit endpoints. 2) shared circuit, is part of our novel approach. It can be used also by flows that are transmitted through ToR switches connected to the circuit endpoints, but originate and/or end at other ToRs. Moreover, the orchestrator may dynamically decide to configure private or shared optical circuits, according to various criteria including current network utilization, traffic flows nature, tenants SLAs, etc. Configuring or changing the optical circuit type requires installing a single OpenFlow rule for each ToR connected to the circuit endpoints; hence, enabling low overhead and fast network configuration. To assess the benefit of such optical circuit configurations, we implement the proposed algorithms and test them over an emulated data and control plane environment. We evaluate the network performance for various network traffic scenarios for both private and shared optical circuits, and compare them to an EPS-only baseline topology with the same total link bandwidth. Our preliminary results show that the shared optical circuits introduce an improvement of 5% to 10% as compared to the commonly used private circuits. The research is partially supported by the European Communitys Seventh Framework Programme (FP7/2001-2013) under grant agreement no. 619572 (COSIGN Project).","PeriodicalId":20607,"journal":{"name":"Proceedings of the 9th ACM International on Systems and Storage Conference","volume":"52 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2016-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Utilizing Optical Circuits in Hybrid Packet/Circuit Data-Center Networks\",\"authors\":\"Y. Ben-Itzhak, C. Caba, José Soler\",\"doi\":\"10.1145/2928275.2933284\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Existing Data Center Networks (DCNs) continue to evolve to keep up with application requirements in terms of bandwidth, latency, agility, etc. According to the updated release of the Cisco Global Cloud Index [1], by 2019, more than 86% of traffic workloads will be processed by cloud DCs. Traditional DCNs, which are based on electrical packet switching (EPS) with hierarchical, tree-like topologies can no longer support future cloud traffic requirements in terms of dynamicity, bandwidth and latency. Hence, existing DCNs can be enhanced with OCS (Optical Circuit Switching), which provides high bandwidth, low latency and low power consumption [2], giving rise to hybrid OCS-EPS topologies. In this research, we assess a virtualized, hybrid, flat DCN topology consisting of a single layer of high radix ToR (Top of Rack) switches, interconnected with each other and through an OCS plane. The benefit of such flat topology is twofold: 1) In terms of bandwidth, over-subscription is reduced, and bisection bandwidth is increased; and 2) In terms of latency, the diameter (longest path) of topology is reduced. Moreover, we present new algorithms and orchestration functionality to detect and offload suitable flows (e.g. elephant flows) from the EPS to the OCS plane. Our DC architecture consists of hybrid EPS-OCS DCN, an Openflow(OF) based control plane, and an orchestration layer. Our orchestration layer decouples the elephant flows detection from the rerouting decision logic in the DCN. Specifically, the elephant flows detection is done by flow tagging in the hypervisor, while the flow rerouting is executed at the EPSs, which are connected directly to the OCS. Hence, it provides a more efficient, scalable, and easy to configure architecture as compared to existing hybrid solutions. The orchestrator monitors the ToR switches by sFlow and detects high volume traffic between two ToRs, exceeding a given bandwidth threshold. Such traffic may consist of either few elephant flows or many mice flows. To further increase the optical circuit utilization, we introduce two types of optical circuits: 1) private circuit, presented in existing solutions, is utilized only by flows that originate and end at the ToR switches connected to the circuit endpoints. 2) shared circuit, is part of our novel approach. It can be used also by flows that are transmitted through ToR switches connected to the circuit endpoints, but originate and/or end at other ToRs. Moreover, the orchestrator may dynamically decide to configure private or shared optical circuits, according to various criteria including current network utilization, traffic flows nature, tenants SLAs, etc. Configuring or changing the optical circuit type requires installing a single OpenFlow rule for each ToR connected to the circuit endpoints; hence, enabling low overhead and fast network configuration. To assess the benefit of such optical circuit configurations, we implement the proposed algorithms and test them over an emulated data and control plane environment. We evaluate the network performance for various network traffic scenarios for both private and shared optical circuits, and compare them to an EPS-only baseline topology with the same total link bandwidth. Our preliminary results show that the shared optical circuits introduce an improvement of 5% to 10% as compared to the commonly used private circuits. The research is partially supported by the European Communitys Seventh Framework Programme (FP7/2001-2013) under grant agreement no. 619572 (COSIGN Project).\",\"PeriodicalId\":20607,\"journal\":{\"name\":\"Proceedings of the 9th ACM International on Systems and Storage Conference\",\"volume\":\"52 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2016-06-06\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Proceedings of the 9th ACM International on Systems and Storage Conference\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1145/2928275.2933284\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of the 9th ACM International on Systems and Storage Conference","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1145/2928275.2933284","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

摘要

现有的数据中心网络(dcn)不断发展,以满足应用程序在带宽、延迟、敏捷性等方面的需求。根据最新发布的思科全球云指数[1],到2019年,超过86%的流量工作负载将由云数据中心处理。传统的dcn是基于电子分组交换(EPS)的分层树状拓扑结构,在动态性、带宽和延迟方面已经不能满足未来云流量的需求。因此,现有的DCNs可以通过OCS(光电路交换)来增强,OCS(光电路交换)提供高带宽,低延迟和低功耗,从而产生混合OCS- eps拓扑。在本研究中,我们评估了一种虚拟化、混合、扁平的DCN拓扑结构,该拓扑结构由单层高基数ToR(机架顶部)交换机组成,通过OCS平面相互连接。这种平面拓扑的好处是双重的:1)在带宽方面,减少了过度订阅,增加了二分带宽;2)在延迟方面,减小了拓扑的直径(最长路径)。此外,我们提出了新的算法和编排功能,以检测和卸载合适的流(例如大象流)从EPS到OCS平面。我们的数据中心架构由混合EPS-OCS DCN、基于Openflow(of)的控制平面和编排层组成。我们的编排层将大象流检测与DCN中的重路由决策逻辑解耦。具体来说,大象流检测是通过管理程序中的流标记完成的,而流重路由是在直接连接到OCS的eps上执行的。因此,与现有的混合解决方案相比,它提供了更高效、可伸缩和易于配置的体系结构。编排器通过sFlow监视ToR交换机,并检测两个ToR之间超过给定带宽阈值的高容量流量。这种交通可能由少量大象流或许多老鼠流组成。为了进一步提高光电路的利用率,我们引入了两种类型的光电路:1)在现有解决方案中提出的专用电路,仅由连接到电路端点的ToR交换机发起和结束的流使用。2)共享电路,是我们新方法的一部分。它也可以用于通过连接到电路端点的ToR交换机传输的流,但起源于和/或结束于其他ToR。此外,编排器可以根据各种标准(包括当前网络利用率、流量流性质、租户sla等)动态决定配置私有或共享光电路。配置或更改光学电路类型需要为连接到电路端点的每个ToR安装单个OpenFlow规则;因此,支持低开销和快速的网络配置。为了评估这种光电路配置的好处,我们实现了所提出的算法,并在模拟数据和控制平面环境中对它们进行了测试。我们评估了专用和共享光电路的各种网络流量场景下的网络性能,并将它们与具有相同总链路带宽的仅eps基线拓扑进行了比较。我们的初步结果表明,与常用的专用电路相比,共享光电路的性能提高了5%至10%。本研究部分由欧共体第七框架计划(FP7/2001-2013)资助,资助协议号:619572 (COSIGN项目)。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
Utilizing Optical Circuits in Hybrid Packet/Circuit Data-Center Networks
Existing Data Center Networks (DCNs) continue to evolve to keep up with application requirements in terms of bandwidth, latency, agility, etc. According to the updated release of the Cisco Global Cloud Index [1], by 2019, more than 86% of traffic workloads will be processed by cloud DCs. Traditional DCNs, which are based on electrical packet switching (EPS) with hierarchical, tree-like topologies can no longer support future cloud traffic requirements in terms of dynamicity, bandwidth and latency. Hence, existing DCNs can be enhanced with OCS (Optical Circuit Switching), which provides high bandwidth, low latency and low power consumption [2], giving rise to hybrid OCS-EPS topologies. In this research, we assess a virtualized, hybrid, flat DCN topology consisting of a single layer of high radix ToR (Top of Rack) switches, interconnected with each other and through an OCS plane. The benefit of such flat topology is twofold: 1) In terms of bandwidth, over-subscription is reduced, and bisection bandwidth is increased; and 2) In terms of latency, the diameter (longest path) of topology is reduced. Moreover, we present new algorithms and orchestration functionality to detect and offload suitable flows (e.g. elephant flows) from the EPS to the OCS plane. Our DC architecture consists of hybrid EPS-OCS DCN, an Openflow(OF) based control plane, and an orchestration layer. Our orchestration layer decouples the elephant flows detection from the rerouting decision logic in the DCN. Specifically, the elephant flows detection is done by flow tagging in the hypervisor, while the flow rerouting is executed at the EPSs, which are connected directly to the OCS. Hence, it provides a more efficient, scalable, and easy to configure architecture as compared to existing hybrid solutions. The orchestrator monitors the ToR switches by sFlow and detects high volume traffic between two ToRs, exceeding a given bandwidth threshold. Such traffic may consist of either few elephant flows or many mice flows. To further increase the optical circuit utilization, we introduce two types of optical circuits: 1) private circuit, presented in existing solutions, is utilized only by flows that originate and end at the ToR switches connected to the circuit endpoints. 2) shared circuit, is part of our novel approach. It can be used also by flows that are transmitted through ToR switches connected to the circuit endpoints, but originate and/or end at other ToRs. Moreover, the orchestrator may dynamically decide to configure private or shared optical circuits, according to various criteria including current network utilization, traffic flows nature, tenants SLAs, etc. Configuring or changing the optical circuit type requires installing a single OpenFlow rule for each ToR connected to the circuit endpoints; hence, enabling low overhead and fast network configuration. To assess the benefit of such optical circuit configurations, we implement the proposed algorithms and test them over an emulated data and control plane environment. We evaluate the network performance for various network traffic scenarios for both private and shared optical circuits, and compare them to an EPS-only baseline topology with the same total link bandwidth. Our preliminary results show that the shared optical circuits introduce an improvement of 5% to 10% as compared to the commonly used private circuits. The research is partially supported by the European Communitys Seventh Framework Programme (FP7/2001-2013) under grant agreement no. 619572 (COSIGN Project).
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信