In this paper, we introduce MMPTCP, a novel transport protocol which aims at unifying the way data is transported in data centres. MMPTCP runs in two phases; initially, it randomly scatters packets in the network under a single congestion window exploiting all available paths. This is beneficial to latency-sensitive flows. During the second phase, MMPTCP runs in Multi-Path TCP (MPTCP) mode, which has been shown to be very efficient for long flows. Initial evaluation shows that our approach significantly improves short flow completion times while providing high throughput for long flows and high overall network utilisation.
{"title":"Short vs. Long Flows: A Battle That Both Can Win","authors":"Morteza Kheirkhah, I. Wakeman, G. Parisis","doi":"10.1145/2785956.2790018","DOIUrl":"https://doi.org/10.1145/2785956.2790018","url":null,"abstract":"In this paper, we introduce MMPTCP, a novel transport protocol which aims at unifying the way data is transported in data centres. MMPTCP runs in two phases; initially, it randomly scatters packets in the network under a single congestion window exploiting all available paths. This is beneficial to latency-sensitive flows. During the second phase, MMPTCP runs in Multi-Path TCP (MPTCP) mode, which has been shown to be very efficient for long flows. Initial evaluation shows that our approach significantly improves short flow completion times while providing high throughput for long flows and high overall network utilisation.","PeriodicalId":268472,"journal":{"name":"Proceedings of the 2015 ACM Conference on Special Interest Group on Data Communication","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123988119","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}
Hitesh Ballani, Paolo Costa, C. Gkantsidis, Matthew P. Grosvenor, T. Karagiannis, Lazaros Koromilas, G. O'Shea
Many network functions executed in modern datacenters, e.g., load balancing, application-level QoS, and congestion control, exhibit three common properties at the data-plane: they need to access and modify state, to perform computations, and to access application semantics -- this is critical since many network functions are best expressed in terms of application-level messages. In this paper, we argue that the end hosts are a natural enforcement point for these functions and we present Eden, an architecture for implementing network functions at datacenter end hosts with minimal network support. Eden comprises three components, a centralized controller, an enclave at each end host, and Eden-compliant applications called stages. To implement network functions, the controller configures stages to classify their data into messages and the enclaves to apply action functions based on a packet's class. Our Eden prototype includes enclaves implemented both in the OS kernel and on programmable NICs. Through case studies, we show how application-level classification and the ability to run actual programs on the data-path allows Eden to efficiently support a broad range of network functions at the network's edge.
{"title":"Enabling End-Host Network Functions","authors":"Hitesh Ballani, Paolo Costa, C. Gkantsidis, Matthew P. Grosvenor, T. Karagiannis, Lazaros Koromilas, G. O'Shea","doi":"10.1145/2785956.2787493","DOIUrl":"https://doi.org/10.1145/2785956.2787493","url":null,"abstract":"Many network functions executed in modern datacenters, e.g., load balancing, application-level QoS, and congestion control, exhibit three common properties at the data-plane: they need to access and modify state, to perform computations, and to access application semantics -- this is critical since many network functions are best expressed in terms of application-level messages. In this paper, we argue that the end hosts are a natural enforcement point for these functions and we present Eden, an architecture for implementing network functions at datacenter end hosts with minimal network support. Eden comprises three components, a centralized controller, an enclave at each end host, and Eden-compliant applications called stages. To implement network functions, the controller configures stages to classify their data into messages and the enclaves to apply action functions based on a packet's class. Our Eden prototype includes enclaves implemented both in the OS kernel and on programmable NICs. Through case studies, we show how application-level classification and the ability to run actual programs on the data-path allows Eden to efficiently support a broad range of network functions at the network's edge.","PeriodicalId":268472,"journal":{"name":"Proceedings of the 2015 ACM Conference on Special Interest Group on Data Communication","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129058065","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}
Software tables are much faster to be updated than hardware ones A hybrid architecture can exploit the following logic: • Forwarding tables updates always happen in software first • Entries are eventually moved to the hardware forwarding tables, offloading the Software forwarder Overlapping entries may break the forwarding decisions in respect to their priorities Deleting entries from hardware tables is usually much faster than adding them: • Installation in hardware tables may require a reorganization of the previously installed Flow Table Entries A hybrid architecture can translate an entry installation into a mix of installation in software tables and deletion from hardware tables
{"title":"Towards Scalable SDN Switches: Enabling Faster Flow Table Entries Installation","authors":"R. Bifulco, A. Matsiuk","doi":"10.1145/2785956.2790008","DOIUrl":"https://doi.org/10.1145/2785956.2790008","url":null,"abstract":" Software tables are much faster to be updated than hardware ones A hybrid architecture can exploit the following logic: • Forwarding tables updates always happen in software first • Entries are eventually moved to the hardware forwarding tables, offloading the Software forwarder Overlapping entries may break the forwarding decisions in respect to their priorities Deleting entries from hardware tables is usually much faster than adding them: • Installation in hardware tables may require a reorganization of the previously installed Flow Table Entries A hybrid architecture can translate an entry installation into a mix of installation in software tables and deletion from hardware tables","PeriodicalId":268472,"journal":{"name":"Proceedings of the 2015 ACM Conference on Special Interest Group on Data Communication","volume":"113 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116083751","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}
F. Schmidt, O. Hohlfeld, René Glebke, Klaus Wehrle
Increasing network speeds challenge the packet processing performance of networked systems. This can mainly be attributed to processing overhead caused by the split between the kernel-space network stack and user-space applications. To mitigate this overhead, we propose Santa, an application agnostic kernel-level cache of frequent requests. By allowing user-space applications to offload frequent requests to the kernel-space, Santa offers drastic performance improvements and unlocks the speed of kernel-space networking for legacy server software without requiring extensive changes.
{"title":"Santa: Faster Packet Delivery for Commonly Wished Replies","authors":"F. Schmidt, O. Hohlfeld, René Glebke, Klaus Wehrle","doi":"10.1145/2785956.2790014","DOIUrl":"https://doi.org/10.1145/2785956.2790014","url":null,"abstract":"Increasing network speeds challenge the packet processing performance of networked systems. This can mainly be attributed to processing overhead caused by the split between the kernel-space network stack and user-space applications. To mitigate this overhead, we propose Santa, an application agnostic kernel-level cache of frequent requests. By allowing user-space applications to offload frequent requests to the kernel-space, Santa offers drastic performance improvements and unlocks the speed of kernel-space networking for legacy server software without requiring extensive changes.","PeriodicalId":268472,"journal":{"name":"Proceedings of the 2015 ACM Conference on Special Interest Group on Data Communication","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116269685","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}
Chuanxiong Guo, Lihua Yuan, Dong Xiang, Yingnong Dang, Ray Huang, D. Maltz, Zhaoyi Liu, Vin Wang, Bin Pang, Hua Chen, Zhi Lin, Varugis Kurien
Can we get network latency between any two servers at any time in large-scale data center networks? The collected latency data can then be used to address a series of challenges: telling if an application perceived latency issue is caused by the network or not, defining and tracking network service level agreement (SLA), and automatic network troubleshooting. We have developed the Pingmesh system for large-scale data center network latency measurement and analysis to answer the above question affirmatively. Pingmesh has been running in Microsoft data centers for more than four years, and it collects tens of terabytes of latency data per day. Pingmesh is widely used by not only network software developers and engineers, but also application and service developers and operators.
{"title":"Pingmesh: A Large-Scale System for Data Center Network Latency Measurement and Analysis","authors":"Chuanxiong Guo, Lihua Yuan, Dong Xiang, Yingnong Dang, Ray Huang, D. Maltz, Zhaoyi Liu, Vin Wang, Bin Pang, Hua Chen, Zhi Lin, Varugis Kurien","doi":"10.1145/2785956.2787496","DOIUrl":"https://doi.org/10.1145/2785956.2787496","url":null,"abstract":"Can we get network latency between any two servers at any time in large-scale data center networks? The collected latency data can then be used to address a series of challenges: telling if an application perceived latency issue is caused by the network or not, defining and tracking network service level agreement (SLA), and automatic network troubleshooting. We have developed the Pingmesh system for large-scale data center network latency measurement and analysis to answer the above question affirmatively. Pingmesh has been running in Microsoft data centers for more than four years, and it collects tens of terabytes of latency data per day. Pingmesh is widely used by not only network software developers and engineers, but also application and service developers and operators.","PeriodicalId":268472,"journal":{"name":"Proceedings of the 2015 ACM Conference on Special Interest Group on Data Communication","volume":"71 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127164722","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The increasing need for ubiquitous Internet connectivity has led to the deployment of wireless access infrastructure, including public WiFi networks used for mobile data offloading. Despite the density of public WiFi hotspots especially in residential areas, these networks are severely underutilized, with an average of ten percent of active users and an average of one connection per user per month [1]. Since the revenue from the public WiFi infrastructure is much lower than the cellular counterpart, the operators seek opportunities to monetize their WiFi networks. One such opportunity emerges from the intention of mobile application vendors to provide near-ubiquitous Internet access for their clients. In fact, the revenue and any potential business opportunities for mobile application vendors are highly dependent on the users’ connection time. Therefore, both network operators and mobile application vendors have incentives to march together towards a ubiquitous WiFi for their clients. However, the traditional wireless architecture hinders any such opportunity, since it requires clients to subscribe and authenticate to the operator before establishing Internet access. To overcome this limitation, we propose application-centric wireless access, at which Service Providers (SPs) can authenticate and connect their clients through public WiFi networks, free of charge. Application-centric wireless access requires the delegation of user authentication and access from the operator to the Service Provider (SP) in a secure and auditable manner. In particular, user authentication requests should be redirected to authentication servers deployed by SPs, while access control should be enabled such that only
{"title":"FreeSurf: Application-Centric Wireless Access with SDN","authors":"Zhen Cao, J. Fitschen, Panagiotis Papadimitriou","doi":"10.1145/2785956.2790000","DOIUrl":"https://doi.org/10.1145/2785956.2790000","url":null,"abstract":"The increasing need for ubiquitous Internet connectivity has led to the deployment of wireless access infrastructure, including public WiFi networks used for mobile data offloading. Despite the density of public WiFi hotspots especially in residential areas, these networks are severely underutilized, with an average of ten percent of active users and an average of one connection per user per month [1]. Since the revenue from the public WiFi infrastructure is much lower than the cellular counterpart, the operators seek opportunities to monetize their WiFi networks. One such opportunity emerges from the intention of mobile application vendors to provide near-ubiquitous Internet access for their clients. In fact, the revenue and any potential business opportunities for mobile application vendors are highly dependent on the users’ connection time. Therefore, both network operators and mobile application vendors have incentives to march together towards a ubiquitous WiFi for their clients. However, the traditional wireless architecture hinders any such opportunity, since it requires clients to subscribe and authenticate to the operator before establishing Internet access. To overcome this limitation, we propose application-centric wireless access, at which Service Providers (SPs) can authenticate and connect their clients through public WiFi networks, free of charge. Application-centric wireless access requires the delegation of user authentication and access from the operator to the Service Provider (SP) in a secure and auditable manner. In particular, user authentication requests should be redirected to authentication servers deployed by SPs, while access control should be enabled such that only","PeriodicalId":268472,"journal":{"name":"Proceedings of the 2015 ACM Conference on Special Interest Group on Data Communication","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127568647","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}
We propose eSDN; a practical approach for deploying new datacenter transports without requiring any changes to the switches. eSDN uses light-weight SDN controllers at the end-hosts for querying network state. It obviates the need for statistics collection by a centralized controller especially on short timescales. We show that eSDN can scale well and allow a range of datacenter transports to be realized.
{"title":"eSDN: Rethinking Datacenter Transports Using End-Host SDN Controllers","authors":"Hasnain Ali Pirzada, M. Mahboob, I. Qazi","doi":"10.1145/2785956.2790022","DOIUrl":"https://doi.org/10.1145/2785956.2790022","url":null,"abstract":"We propose eSDN; a practical approach for deploying new datacenter transports without requiring any changes to the switches. eSDN uses light-weight SDN controllers at the end-hosts for querying network state. It obviates the need for statistics collection by a centralized controller especially on short timescales. We show that eSDN can scale well and allow a range of datacenter transports to be realized.","PeriodicalId":268472,"journal":{"name":"Proceedings of the 2015 ACM Conference on Special Interest Group on Data Communication","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125976333","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}
Distributed consensus is fundamental in distributed systems for achieving fault-tolerance. The Paxos algorithm has long dominated this domain, although it has been recently challenged by algorithms such as Raft and Viewstamped Replication Revisited. These algorithms rely on Paxos's original assumptions, unfortunately these assumptions are now at odds with the reality of the modern internet. Our insight is that current consensus algorithms have significant availability issues when deployed outside the well defined context of the datacenter. To illustrate this problem, we developed Coracle, a tool for evaluating distributed consensus algorithms in settings that more accurately represent realistic deployments. We have used Coracle to test two examples of network configurations that contradict the liveness claims of the Raft algorithm. Through the process of exercising these algorithms under more realistic assumptions, we demonstrate wider availability issues faced by consensus algorithms when deployed on real world networks.
{"title":"Coracle: Evaluating Consensus at the Internet Edge","authors":"H. Howard, J. Crowcroft","doi":"10.1145/2785956.2790010","DOIUrl":"https://doi.org/10.1145/2785956.2790010","url":null,"abstract":"Distributed consensus is fundamental in distributed systems for achieving fault-tolerance. The Paxos algorithm has long dominated this domain, although it has been recently challenged by algorithms such as Raft and Viewstamped Replication Revisited. These algorithms rely on Paxos's original assumptions, unfortunately these assumptions are now at odds with the reality of the modern internet. Our insight is that current consensus algorithms have significant availability issues when deployed outside the well defined context of the datacenter. To illustrate this problem, we developed Coracle, a tool for evaluating distributed consensus algorithms in settings that more accurately represent realistic deployments. We have used Coracle to test two examples of network configurations that contradict the liveness claims of the Raft algorithm. Through the process of exercising these algorithms under more realistic assumptions, we demonstrate wider availability issues faced by consensus algorithms when deployed on real world networks.","PeriodicalId":268472,"journal":{"name":"Proceedings of the 2015 ACM Conference on Special Interest Group on Data Communication","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126687330","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}
Brandon Schlinker, Radhika Niranjan Mysore, Sean Smith, J. Mogul, Amin Vahdat, Minlan Yu, Ethan Katz-Bassett, Michael Rubin
The design space for large, multipath datacenter networks is large and complex, and no one design fits all purposes. Network architects must trade off many criteria to design cost-effective, reliable, and maintainable networks, and typically cannot explore much of the design space. We present Condor, our approach to enabling a rapid, efficient design cycle. Condor allows architects to express their requirements as constraints via a Topology Description Language (TDL), rather than having to directly specify network structures. Condor then uses constraint-based synthesis to rapidly generate candidate topologies, which can be analyzed against multiple criteria. We show that TDL supports concise descriptions of topologies such as fat-trees, BCube, and DCell; that we can generate known and novel variants of fat-trees with simple changes to a TDL file; and that we can synthesize large topologies in tens of seconds. We also show that Condor supports the daunting task of designing multi-phase network expansions that can be carried out on live networks.
{"title":"Condor: Better Topologies Through Declarative Design","authors":"Brandon Schlinker, Radhika Niranjan Mysore, Sean Smith, J. Mogul, Amin Vahdat, Minlan Yu, Ethan Katz-Bassett, Michael Rubin","doi":"10.1145/2785956.2787476","DOIUrl":"https://doi.org/10.1145/2785956.2787476","url":null,"abstract":"The design space for large, multipath datacenter networks is large and complex, and no one design fits all purposes. Network architects must trade off many criteria to design cost-effective, reliable, and maintainable networks, and typically cannot explore much of the design space. We present Condor, our approach to enabling a rapid, efficient design cycle. Condor allows architects to express their requirements as constraints via a Topology Description Language (TDL), rather than having to directly specify network structures. Condor then uses constraint-based synthesis to rapidly generate candidate topologies, which can be analyzed against multiple criteria. We show that TDL supports concise descriptions of topologies such as fat-trees, BCube, and DCell; that we can generate known and novel variants of fat-trees with simple changes to a TDL file; and that we can synthesize large topologies in tens of seconds. We also show that Condor supports the daunting task of designing multi-phase network expansions that can be carried out on live networks.","PeriodicalId":268472,"journal":{"name":"Proceedings of the 2015 ACM Conference on Special Interest Group on Data Communication","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126593758","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}
Content Delivery Networks (CDNs) deliver much of the world's web, video, and application content on the Internet today. A key component of a CDN is the mapping system that uses the DNS protocol to route each client's request to a ``proximal'' server that serves the requested content. While traditional mapping systems identify a client using the IP of its name server, we describe our experience in building and rolling-out a novel system called end-user mapping that identifies the client directly by using a prefix of the client's IP address. Using measurements from Akamai's production network during the roll-out, we show that end-user mapping provides significant performance benefits for clients who use public resolvers, including an eight-fold decrease in mapping distance, a two-fold decrease in RTT and content download time, and a 30% improvement in the time-to-first byte. We also quantify the scaling challenges in implementing end-user mapping such as the 8-fold increase in DNS queries. Finally, we show that a CDN with a larger number of deployment locations is likely to benefit more from end-user mapping than a CDN with a smaller number of deployments.
{"title":"End-User Mapping: Next Generation Request Routing for Content Delivery","authors":"Fangfei Chen, R. Sitaraman, Marcelo Torres","doi":"10.1145/2785956.2787500","DOIUrl":"https://doi.org/10.1145/2785956.2787500","url":null,"abstract":"Content Delivery Networks (CDNs) deliver much of the world's web, video, and application content on the Internet today. A key component of a CDN is the mapping system that uses the DNS protocol to route each client's request to a ``proximal'' server that serves the requested content. While traditional mapping systems identify a client using the IP of its name server, we describe our experience in building and rolling-out a novel system called end-user mapping that identifies the client directly by using a prefix of the client's IP address. Using measurements from Akamai's production network during the roll-out, we show that end-user mapping provides significant performance benefits for clients who use public resolvers, including an eight-fold decrease in mapping distance, a two-fold decrease in RTT and content download time, and a 30% improvement in the time-to-first byte. We also quantify the scaling challenges in implementing end-user mapping such as the 8-fold increase in DNS queries. Finally, we show that a CDN with a larger number of deployment locations is likely to benefit more from end-user mapping than a CDN with a smaller number of deployments.","PeriodicalId":268472,"journal":{"name":"Proceedings of the 2015 ACM Conference on Special Interest Group on Data Communication","volume":"387 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131935484","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：