Although distributed key generation (DKG) has been studied for some time, it has never been examined outside of the synchronous setting. We present the first realistic DKG architecture for use over the Internet. We propose a practical system model and define an efficient verifiable secret sharing scheme in it. We observe the necessity of Byzantine agreement for asynchronous DKG and analyze the difficulty of using a randomized protocol for it. Using our verifiable secret sharing scheme and a leader-based agreement protocol, we then design a DKG protocol for public-key cryptography. Finally, along with traditional proactive security, we also introduce group modification primitives in our system.
{"title":"Distributed Key Generation for the Internet","authors":"Aniket Kate, I. Goldberg","doi":"10.1109/ICDCS.2009.21","DOIUrl":"https://doi.org/10.1109/ICDCS.2009.21","url":null,"abstract":"Although distributed key generation (DKG) has been studied for some time, it has never been examined outside of the synchronous setting. We present the first realistic DKG architecture for use over the Internet. We propose a practical system model and define an efficient verifiable secret sharing scheme in it. We observe the necessity of Byzantine agreement for asynchronous DKG and analyze the difficulty of using a randomized protocol for it. Using our verifiable secret sharing scheme and a leader-based agreement protocol, we then design a DKG protocol for public-key cryptography. Finally, along with traditional proactive security, we also introduce group modification primitives in our system.","PeriodicalId":387968,"journal":{"name":"2009 29th IEEE International Conference on Distributed Computing Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2009-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122498834","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}
A strongly fair schedule is one in which tasks that are enabled infinitely often are also executed infinitely often. When tasks execute atomically, a strongly fair scheduler can be implemented in a maximal manner. That is, an algorithm exists that, for any valid schedule, is capable of generating that schedule. We show that this assumption of atomicity is necessary. That is, when task execution can be interleaved with other tasks, no algorithm is capable of generating all valid schedules. In other words, any algorithm that correctly generates some strongly fair schedules must also be incapable of generating some other valid schedules. This impossibility result is the first example of an implementable UNITY specification for which no maximal solution exists.
{"title":"On the Impossibility of Maximal Scheduling for Strong Fairness with Interleaving","authors":"Matthew Lang, P. Sivilotti","doi":"10.1109/ICDCS.2009.59","DOIUrl":"https://doi.org/10.1109/ICDCS.2009.59","url":null,"abstract":"A strongly fair schedule is one in which tasks that are enabled infinitely often are also executed infinitely often. When tasks execute atomically, a strongly fair scheduler can be implemented in a maximal manner. That is, an algorithm exists that, for any valid schedule, is capable of generating that schedule. We show that this assumption of atomicity is necessary. That is, when task execution can be interleaved with other tasks, no algorithm is capable of generating all valid schedules. In other words, any algorithm that correctly generates some strongly fair schedules must also be incapable of generating some other valid schedules. This impossibility result is the first example of an implementable UNITY specification for which no maximal solution exists.","PeriodicalId":387968,"journal":{"name":"2009 29th IEEE International Conference on Distributed Computing Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2009-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122800197","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 email volume per mailbox has largely remained low and unchanged in the past several decades, and hence mail server performance has largely remained a secondary issue. The steep rise in the amount of unsolicited emails, i.e. spam, in the past decade, however, has permanently disrupted this tranquility of largely steady email volume and turned mail server performance into an increasingly important issue. In this paper, we point out that modern mail servers were not originally designed with email spam in mind, and as such, as the "common case'' workload for mail servers has shifted from legitimate emails to spam emails, we argue it is time to revisit mail server architecture design in following the system design principle of optimizing the common case. In particular, we show how to optimize the performance of three major components of modern mail servers, the concurrency architecture, the disk I/O, and DNSBL lookups, by exploiting the new "common case" workload. An evaluation of our prototype implementation of the enhanced postfix architecture shows that the optimizations significantly reduce the CPU, disk, and network resource consumptions, and improves the throughput of the mail server by 18% under a university departmental mail server workload and by 40% under a spam sinkhole workload.
{"title":"The Case for Spam-Aware High Performance Mail Server Architecture","authors":"Abhinav Pathak, Syed Ali, Raza Jafri, Y. C. Hu","doi":"10.1109/ICDCS.2009.74","DOIUrl":"https://doi.org/10.1109/ICDCS.2009.74","url":null,"abstract":"The email volume per mailbox has largely remained low and unchanged in the past several decades, and hence mail server performance has largely remained a secondary issue. The steep rise in the amount of unsolicited emails, i.e. spam, in the past decade, however, has permanently disrupted this tranquility of largely steady email volume and turned mail server performance into an increasingly important issue. In this paper, we point out that modern mail servers were not originally designed with email spam in mind, and as such, as the \"common case'' workload for mail servers has shifted from legitimate emails to spam emails, we argue it is time to revisit mail server architecture design in following the system design principle of optimizing the common case. In particular, we show how to optimize the performance of three major components of modern mail servers, the concurrency architecture, the disk I/O, and DNSBL lookups, by exploiting the new \"common case\" workload. An evaluation of our prototype implementation of the enhanced postfix architecture shows that the optimizations significantly reduce the CPU, disk, and network resource consumptions, and improves the throughput of the mail server by 18% under a university departmental mail server workload and by 40% under a spam sinkhole workload.","PeriodicalId":387968,"journal":{"name":"2009 29th IEEE International Conference on Distributed Computing Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2009-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134176510","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}
Concurrency control is a core component in optimistic replication systems. To detect concurrent updates, the system associates each replicated object with metadata, such as, version vectors or causal graphs exchanged on synchronization opportunities. However, the size of such metadata increases at least linearly with the number of active sites. With recent trends in cloud computing, multi-regional collaboration, and mobile networks, the number of sites within a single replication system becomes very large. This imposes substantial overhead in communication and computation on every site. In this paper, we first present three version vector implementations that significantly reduce the cost of vector exchange by incrementally transferring vector elements. Basic rotating vectors (BRV) support systems providing no conflict reconciliation, whereas conflict rotating vectors (CRV) extend BRV to overcome this limitation. Skip rotating vectors (SRV) based on CRV further reduce data transmission. We show that both BRV and SRV are optimal implementations of version vectors, which, in turn, have minimal storage complexity among all known concurrency control schemes for state-transfer systems. We then present a causal graph exchange algorithm for operation-transfer systems with optimal communication overhead. All these algorithms adopt network pipelining to reduce running time.
{"title":"On Optimal Concurrency Control for Optimistic Replication","authors":"Weihan Wang, C. Amza","doi":"10.1109/ICDCS.2009.71","DOIUrl":"https://doi.org/10.1109/ICDCS.2009.71","url":null,"abstract":"Concurrency control is a core component in optimistic replication systems. To detect concurrent updates, the system associates each replicated object with metadata, such as, version vectors or causal graphs exchanged on synchronization opportunities. However, the size of such metadata increases at least linearly with the number of active sites. With recent trends in cloud computing, multi-regional collaboration, and mobile networks, the number of sites within a single replication system becomes very large. This imposes substantial overhead in communication and computation on every site. In this paper, we first present three version vector implementations that significantly reduce the cost of vector exchange by incrementally transferring vector elements. Basic rotating vectors (BRV) support systems providing no conflict reconciliation, whereas conflict rotating vectors (CRV) extend BRV to overcome this limitation. Skip rotating vectors (SRV) based on CRV further reduce data transmission. We show that both BRV and SRV are optimal implementations of version vectors, which, in turn, have minimal storage complexity among all known concurrency control schemes for state-transfer systems. We then present a causal graph exchange algorithm for operation-transfer systems with optimal communication overhead. All these algorithms adopt network pipelining to reduce running time.","PeriodicalId":387968,"journal":{"name":"2009 29th IEEE International Conference on Distributed Computing Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2009-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134628005","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}
Aggregation operations are important for users to get summarization information in WSN applications. As large numbers of applications only require approximate aggregation results rather than the exact ones, some approximate aggregation algorithms are proposed to save energy. However, the error bounds of these algorithms are fixed and it is impossible to adjust their error bounds automatically. Therefore, these algorithms cannot reach arbitrary precision requirement given by user. This paper proposes a sampling based approximate aggregation algorithm to satisfy the requirement of arbitrary precision. Besides, two sample data adaptive algorithms are also provided. One is to adapt the sample with the varying of precision requirement. The other is to adapt the sample with the varying of the sensed data in networks. The theoretical analysis and experiment results show that the proposed algorithms have high performance in terms of accuracy and energy cost.
{"title":"Sampling Based (epsilon, delta)-Approximate Aggregation Algorithm in Sensor Networks","authors":"Siyao Cheng, Jianzhong Li","doi":"10.1109/ICDCS.2009.8","DOIUrl":"https://doi.org/10.1109/ICDCS.2009.8","url":null,"abstract":"Aggregation operations are important for users to get summarization information in WSN applications. As large numbers of applications only require approximate aggregation results rather than the exact ones, some approximate aggregation algorithms are proposed to save energy. However, the error bounds of these algorithms are fixed and it is impossible to adjust their error bounds automatically. Therefore, these algorithms cannot reach arbitrary precision requirement given by user. This paper proposes a sampling based approximate aggregation algorithm to satisfy the requirement of arbitrary precision. Besides, two sample data adaptive algorithms are also provided. One is to adapt the sample with the varying of precision requirement. The other is to adapt the sample with the varying of the sensed data in networks. The theoretical analysis and experiment results show that the proposed algorithms have high performance in terms of accuracy and energy cost.","PeriodicalId":387968,"journal":{"name":"2009 29th IEEE International Conference on Distributed Computing Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2009-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114479560","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}
TDMA has been proposed as a MAC protocol for wireless sensor networks (WSNs) due to its efficiency in high WSN load. However, TDMA is plagued with shortcomings; we present modifications to TDMA that will allow for the same efficiency of TDMA, while allowing the network to conserve energy during times of low load (when there is no activity being detected). Recognizing that aggregation plays an essential role in WSNs, TDMA-ASAP adds to TDMA: (a) transmission parallelism based on a level-by-level localized graph-coloring, (b) appropriate sleeping between transmissions ("napping"), (c) judicious and controlled TDMA slot stealing to avoid empty slots to be unused and (d) intelligent scheduling/ordering transmissions. Our results show that TDMA-ASAP's unique combination of TDMA, slot-stealing, napping, and message aggregation significantly outperforms other hybrid WSN MAC algorithms and has a performance that is close to optimal in terms of energy consumption and overall delay.
{"title":"TDMA-ASAP: Sensor Network TDMA Scheduling with Adaptive Slot-Stealing and Parallelism","authors":"Sameh Gobriel, D. Mossé, Robert Cleric","doi":"10.1109/ICDCS.2009.80","DOIUrl":"https://doi.org/10.1109/ICDCS.2009.80","url":null,"abstract":"TDMA has been proposed as a MAC protocol for wireless sensor networks (WSNs) due to its efficiency in high WSN load. However, TDMA is plagued with shortcomings; we present modifications to TDMA that will allow for the same efficiency of TDMA, while allowing the network to conserve energy during times of low load (when there is no activity being detected). Recognizing that aggregation plays an essential role in WSNs, TDMA-ASAP adds to TDMA: (a) transmission parallelism based on a level-by-level localized graph-coloring, (b) appropriate sleeping between transmissions (\"napping\"), (c) judicious and controlled TDMA slot stealing to avoid empty slots to be unused and (d) intelligent scheduling/ordering transmissions. Our results show that TDMA-ASAP's unique combination of TDMA, slot-stealing, napping, and message aggregation significantly outperforms other hybrid WSN MAC algorithms and has a performance that is close to optimal in terms of energy consumption and overall delay.","PeriodicalId":387968,"journal":{"name":"2009 29th IEEE International Conference on Distributed Computing Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2009-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124140613","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 consider the distributed complexity of the stable marriage problem. In this problem, the communication graph is undirected and bipartite, and each node ranks its neighbors. Given a matching of the nodes, a pair of nodes is called blocking if they prefer each other to their assigned match. A matching is called stable if it does not induce any blocking pair. In the distributed model, nodes exchange messages in each round over the communication links, until they find a stable matching. We show that if messages may contain at most B bits each, then any distributed algorithm that solves the stable marriage problem requires Omega(sqrt(n/(B log n))) communication rounds in the worst case, even for graphs of diameter Theta (log n), where n is the number of nodes in the graph. Furthermore, the lower bound holds even if we allow the output to contain O(sqrt(n)) blocking pairs. We also consider epsilon-stability, where a pair is called epsilon-blocking if they can improve the quality of their match by more than an epsilon fraction, for some 0
{"title":"A Note on Distributed Stable Matching","authors":"Alexander Kipnis, B. Patt-Shamir","doi":"10.1109/ICDCS.2009.69","DOIUrl":"https://doi.org/10.1109/ICDCS.2009.69","url":null,"abstract":"We consider the distributed complexity of the stable marriage problem. In this problem, the communication graph is undirected and bipartite, and each node ranks its neighbors. Given a matching of the nodes, a pair of nodes is called blocking if they prefer each other to their assigned match. A matching is called stable if it does not induce any blocking pair. In the distributed model, nodes exchange messages in each round over the communication links, until they find a stable matching. We show that if messages may contain at most B bits each, then any distributed algorithm that solves the stable marriage problem requires Omega(sqrt(n/(B log n))) communication rounds in the worst case, even for graphs of diameter Theta (log n), where n is the number of nodes in the graph. Furthermore, the lower bound holds even if we allow the output to contain O(sqrt(n)) blocking pairs. We also consider epsilon-stability, where a pair is called epsilon-blocking if they can improve the quality of their match by more than an epsilon fraction, for some 0","PeriodicalId":387968,"journal":{"name":"2009 29th IEEE International Conference on Distributed Computing Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2009-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130472351","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}
R. Baldoni, Silvia Bonomi, Anne-Marie Kermarrec, M. Raynal
Providing distributed processes with concurrent objects is a fundamental service that has to be offered by any distributed system. The classical shared read/write register is one of the most basic ones. Several protocols have been proposed that build an atomic register on top of an asyn- chronous message-passing system prone to process crashes. In the same spirit, this paper addresses the implementation of a regular register (a weakened form of an atomic register) in an asynchronous dynamic message-passing system. The aim is here to cope with the net effect of the adversaries that are asynchrony and dynamicity (the fact that processes can enter and leave the system). The paper focuses on the class of dynamic systems the churn rate c of which is constant. It presents two protocols, one applicable to synchronous dynamic message passing systems, the other one to even- tually synchronous dynamic systems. Both protocols rely on an appropriate broadcast communication service (similar to a reliable broadcast). Each requires a specific constraint on the churn rate c. Both protocols are first presented in an as intuitive as possible way, and are then proved correct.
{"title":"Implementing a Register in a Dynamic Distributed System","authors":"R. Baldoni, Silvia Bonomi, Anne-Marie Kermarrec, M. Raynal","doi":"10.1109/ICDCS.2009.46","DOIUrl":"https://doi.org/10.1109/ICDCS.2009.46","url":null,"abstract":"Providing distributed processes with concurrent objects is a fundamental service that has to be offered by any distributed system. The classical shared read/write register is one of the most basic ones. Several protocols have been proposed that build an atomic register on top of an asyn- chronous message-passing system prone to process crashes. In the same spirit, this paper addresses the implementation of a regular register (a weakened form of an atomic register) in an asynchronous dynamic message-passing system. The aim is here to cope with the net effect of the adversaries that are asynchrony and dynamicity (the fact that processes can enter and leave the system). The paper focuses on the class of dynamic systems the churn rate c of which is constant. It presents two protocols, one applicable to synchronous dynamic message passing systems, the other one to even- tually synchronous dynamic systems. Both protocols rely on an appropriate broadcast communication service (similar to a reliable broadcast). Each requires a specific constraint on the churn rate c. Both protocols are first presented in an as intuitive as possible way, and are then proved correct.","PeriodicalId":387968,"journal":{"name":"2009 29th IEEE International Conference on Distributed Computing Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2009-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133608537","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}
In this paper, we focus on challenges of supporting a decision support system (DSS) based on a hybrid approach (i.e. a federation system with data placement) for agile business intelligence applications. A DSS needs to be designed to handle a workload of potentially complex queries for important decision-making processes. The response time requirement (and a realistic goal) for such a DSS is near real time. The users of a DSS care about not only the response time but also the time stamp of the business operation report since both of them introduce uncertainty and risks to business decision-making. In our proposed DSS, each report is assigned with a business value; denoting its importance to business decision-making. An Information Value (IV) is a business value of a report discounted by time to reflex the uncertainty and risks associated with the computational latency and synchronization latency. We propose a novel Information Value-driven Query Processing (IVQP) framework specific for near real time DSS applications. The framework enables dynamic query plan selection by taking into account of information value and adaptation for online-arrival ad hoc queries. The framework works with single query as well as a workload of queries. The experimental results based on synthetic data and TPC-H show the effectiveness of our approach in achieving optimal information values for the workloads.
{"title":"Information Value-Driven Near Real-Time Decision Support Systems","authors":"Ying Yan, Wen-Syan Li, Jian Xu","doi":"10.1109/ICDCS.2009.17","DOIUrl":"https://doi.org/10.1109/ICDCS.2009.17","url":null,"abstract":"In this paper, we focus on challenges of supporting a decision support system (DSS) based on a hybrid approach (i.e. a federation system with data placement) for agile business intelligence applications. A DSS needs to be designed to handle a workload of potentially complex queries for important decision-making processes. The response time requirement (and a realistic goal) for such a DSS is near real time. The users of a DSS care about not only the response time but also the time stamp of the business operation report since both of them introduce uncertainty and risks to business decision-making. In our proposed DSS, each report is assigned with a business value; denoting its importance to business decision-making. An Information Value (IV) is a business value of a report discounted by time to reflex the uncertainty and risks associated with the computational latency and synchronization latency. We propose a novel Information Value-driven Query Processing (IVQP) framework specific for near real time DSS applications. The framework enables dynamic query plan selection by taking into account of information value and adaptation for online-arrival ad hoc queries. The framework works with single query as well as a workload of queries. The experimental results based on synthetic data and TPC-H show the effectiveness of our approach in achieving optimal information values for the workloads.","PeriodicalId":387968,"journal":{"name":"2009 29th IEEE International Conference on Distributed Computing Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2009-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131141580","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}
With the ever-increasing P2P Internet traffic, recently much attention has been paid to the topology mismatch between the P2P overlay and the underlying network due to the large amount of cross-ISP traffic. Mainly focusing on BitTorrent-like file sharing systems, several recent studies have demonstrated how to efficiently bridge the overlay and the underlying network by leveraging the existing infrastructure, such as CDN services or developing new application-ISP interfaces, such as P4P. However, so far the traffic locality in existing P2P live streaming systems has not been well studied. In this work, taking PPLive as an example, we examine traffic locality in Internet P2P streaming systems. Our measurement results on both popular and unpopular channels from various locations show that current PPLive traffic is highly localized at the ISP level. In particular, we find: (1) a PPLive peer mainly obtains peer lists referred by its connected neighbors (rather than tracker servers) and up to 90% of listed peers are from the same ISP as the requesting peer; (2) the major portion of the streaming traffic received by a requesting peer (up to 88% in popular channels) is served by peers in the same ISP as the requestor; (3) the top 10% of the connected peers provide most (about 70%) of the requested streaming data and these top peers have smaller RTT to the requesting peer. Our study reveals that without using any topology information or demanding any infrastructure support, PPLive achieves such high ISP level traffic locality spontaneously with its decentralized, latency based, neighbor referral peer selection strategy. These findings provide some new insights for better understanding and optimizing the network- and user-level performance in practical P2P live streaming systems.
{"title":"A Case Study of Traffic Locality in Internet P2P Live Streaming Systems","authors":"Yao Liu, Lei Guo, Fei Li, Songqing Chen","doi":"10.1109/ICDCS.2009.50","DOIUrl":"https://doi.org/10.1109/ICDCS.2009.50","url":null,"abstract":"With the ever-increasing P2P Internet traffic, recently much attention has been paid to the topology mismatch between the P2P overlay and the underlying network due to the large amount of cross-ISP traffic. Mainly focusing on BitTorrent-like file sharing systems, several recent studies have demonstrated how to efficiently bridge the overlay and the underlying network by leveraging the existing infrastructure, such as CDN services or developing new application-ISP interfaces, such as P4P. However, so far the traffic locality in existing P2P live streaming systems has not been well studied. In this work, taking PPLive as an example, we examine traffic locality in Internet P2P streaming systems. Our measurement results on both popular and unpopular channels from various locations show that current PPLive traffic is highly localized at the ISP level. In particular, we find: (1) a PPLive peer mainly obtains peer lists referred by its connected neighbors (rather than tracker servers) and up to 90% of listed peers are from the same ISP as the requesting peer; (2) the major portion of the streaming traffic received by a requesting peer (up to 88% in popular channels) is served by peers in the same ISP as the requestor; (3) the top 10% of the connected peers provide most (about 70%) of the requested streaming data and these top peers have smaller RTT to the requesting peer. Our study reveals that without using any topology information or demanding any infrastructure support, PPLive achieves such high ISP level traffic locality spontaneously with its decentralized, latency based, neighbor referral peer selection strategy. These findings provide some new insights for better understanding and optimizing the network- and user-level performance in practical P2P live streaming systems.","PeriodicalId":387968,"journal":{"name":"2009 29th IEEE International Conference on Distributed Computing Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2009-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115531493","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}