Search

Accurate and effective estimation of TCP operating point impacts directly on TCP performance, especially in modern high-speed networks. In this paper, we focus on two crucial questions: 1) In order to guarantee the performance and stability of TCP, how accurately does the TCP operating point need to be estimated? 2) Ahead of the occurrence of network congestion, how accurately can TCP estimate its operating points at the end host? We declare that the estimator which is based on TCP packet series can satisfy the accuracy TCP required.

The accuracy of network coordinate (NC) which is comprehensively applied is suffering seriously from Triangle Inequality Violations (TIVs). A novel approach - Embeddable Overlay Networks (EON) - has been proposed to address this problem. It runs NC on an overlay environment where Round Trip Times (RTTs) are measured with respect to overlay forwarding that has eliminated all the TIVs. In this paper, we conduct a much more comprehensive study by applying the idea of EON on four existing NC systems - GNP, NPS, Phoenix and PIC. Simulations using original data and data of EON are conducted and the results are compared. Through empirical study we discover that the accuracy has been improved significantly on all of the four NC systems according to the three metrics of relative error, distortion and nearest neighbor loss.

Our understanding of IPv6 usage is quite limited. In this paper, we carry out a detailed analysis of Netflow records collected from a China Tier-1 ISP. Many aspects related to traffic attributes and deployment issues are investigated: application types, transitional technologies, interface identifier assignment schemes, etc. Dominating of video streaming and P2P traffic as well as the widely use of random identifiers for privacy are most peculiar characteristics in China's IPv6 networks. To have a better understanding of how IPv6 network is used, we studied users' behavior from several perspectives, such as the traffic volume produced by each user, number of applications he uses and the frequency he uses IPv6 networks. Results show that IPv6 is only effectively used by a small fraction of the users. On the whole, our study suggests that from our vantage point, China's IPv6 network is in the transitional phase from the initial experimental period to practical application.

Recent work on securing the Network Coordinate (NC) service attempts to defend attacks based on nodes' past behavior and activities. They detect malicious nodes only after attacks occurred. In order to secure NC system before malicious nodes mount attacks, we propose a social-link defense system for NC service--Sniper. We are able to provide NC system with trusted neighbors (landmarks) using social-link filtering technique. Our preliminary results demonstrate that Sniper can protect NC system successfully before they are attacked by malicious nodes.

Anycast paradigm has been widely adopted by Internet application to find nearby resources. The current IP anycast implementation suffers from poor scalability. To overcome that, this paper proposes Proxima, a network coordinates (NC) based application layer infrastructure, which is capable to provide lightweight and flexible anycast service. Proxima accurately chooses the best receiver from a group of candidates according to the Round Trip Time (RTT) for certain application. Our experimental results have demonstrated the excellent performance of Proxima in real sever selection scenario.

Broadcatching is a promising mechanism to improve the experience of BitTorrent users by automatically downloading files advertised through RSS feeds. However, though widely used, the mechanism itself has not been well studied. In this paper, we conducted extensive experiments on PlanetLab to evaluate the performance of Broadcatching under different typical scenarios. The results demonstrated the effectiveness of the Broadcatching: it reduces the average completion time and downloading failure ratio. It also improves the overall fairness of the system: the subscribers are encouraged to share more while downloading faster, which results in the increased share ratio. Our study is the first work to systematically evaluate the benefit of Broadcatching and sheds lights on how to improve performance of BitTorrrent by manipulating peer's behavior like Broadcatching.

In recent years, network coordinate systems which map nodes into a geometrical space can effectively support overlay applications relying on topology-awareness. However, these systems base on an ideal assumption that the nodes in them are honest to cooperate with each other. Although there have been some studies about attacks on network coordinate systems, the effect of attacks on PIC---one of the representative systems---has not been studied. Moreover, since PIC itself has proposed a security policy, how well it can protect PIC from attacks is another significant problem to be researched. We apply four typical attacks on PIC with security and without security. Our extensive experiments show that PIC is vulnerable by attacks and when the percentage of malicious nodes is more than 40%, PIC with security performs barely better than without security.

As the substrate of structured peer-to-peer systems, distributed hash table (DHT) plays a key role in P2P routing infrastructures. Traditional DHT does not consider the location of the nodes for the assignment of identifiers, which will result in high end-to-end latency on DHT-based overlay networks. In this paper, we propose a design of locality-aware DHT called LDHT, which exploits network locality on DHT-based systems. Instead of assigning uniform random node identifiers in traditional DHT, nodes in LDHT are assigned locality-aware identifiers according to their autonomous system numbers (ASNs). As a result, each node will have more nearby neighbors than faraway neighbors in the overlay. We evaluate the performance of LDHT on different kinds of typical DHT protocols and on various topologies. The results show that LDHT improves the traditional DHT protocols a lot in terms of end-to-end latency, without introducing additional overhead. It is indicated that LDHT is fit for different kinds of DHT protocols and can work effectively on all structured P2P systems including Chord, Symphony and Kademlia.

A large number of overlay protocols have been developed, almost all of which assume each host has two-way communication capability. However, this does not hold as the deployment of firewalls and Network Address Translators (NAT) is widespread in the current Internet, which is a challenge to the design and implementation of overlay models and protocols. In this paper, we present Potential-based Multicast Tree Algorithm (PMTA) to enhance the multicast tree construction in presence of connectivity restricted hosts. We evaluate PMTA and previous multicast tree protocols based on real Internet end-to-end delay datasets. According to evaluation results, PMTA outperforms those protocols in terms of all metrics. PMTA reduces ARDP by 26%, and it also results in 23%-54% reduction in average overlay latencies. As the results suggest, PMTA can build efficient and effective multicast tree and is suitable for Internet multicast applications in the presence of connectivity restricted hosts.

In this paper, we propose NTS6, an IPv6 based network topology service system, which provides a query and response service of network topology information between any pair of hosts accessing IPv6 Internet via CERNET2. The NTS6 system periodically collects the latest OSPF, BGP and SNMP information of CERNET2. Based on these information, NTS6 provides network topology information such as net distance, available bandwidth and path packet loss in response to users' query. The real-world experiment on CERNET2 confirms the accuracy of our system. Also, NTS6 now performs well on large deployed maze P2P file-sharing system. We believe NTS6 will be a powerful middleware for other applications.

In this paper, we propose Rainbow, a mesh based locality-aware P2P overlay multicast system. Nodes first self-organize into a two-layer overlay. A simple and accurate network coordinate system is used for more reliable node clustering. Pull-based method is used in data packet delivery. According to our simulation, the ARDP of Rainbow ranges from 63% to 76% of that in Binning overlay. Our real-world experiment on PlanetLab shows that the average packet loss rate of Rainbow is about 50% of that in ChainSaw.

Hybrid Overlay Network (HONet) is a locality-aware hybrid overlay architecture, which combines the scalability and simplicity of a structured overlay with the connection flexibility of an unstructured overlay. In this paper, we propose Canicula, an improved HONet. First, using a simple and accurate network coordinate system, Canicula achieves more reliable node clustering. Secondly, Canicula uses an enhanced overlay construction method to reduce the impact of guarded hosts. The experimental results on over 450 PlanetLab nodes show that the ARDP in Canicula is only 45% of that in flat bidirectional Chord.

In practice, because of their saturating affects, potentially all traditional transmission control protocol (TCP) congestion estimators effectively degrade the operational performance and efficiency of virtually all high-speed networks. Here the authors adopt a novel idea of using queueing-based estimator to estimate the TCP operating points ahead of their occurrence of network congestion. For this purpose the authors employ recent work of packet time-series analysis. In order to demonstrate the effectiveness of the idea, the authors embed the new estimator into several TCP variants in NS-2 and Linux kernel. For this evaluation the authors have programmed the simulation system to simulate a wide range of classic operating environments and setup many practical test-bed emulators for measurements. The authors results show that the idea of using a queueing-based estimator provides us far more earlier, sensitive and consistent estimation for the operating point than the classic loss/delay-based TCP traffic estimators. The authors then show that the new approach improves the overall estimation efficiency and provides higher performance and fairness when monitoring TCP traffic.

A Network Coordinate (NC) system is an efficient mechanism to predict Internet distance with scalable measurements. In this paper, we focus on the node churn problem - the continuous process of nodes arrival and departure - in distributed applications. Studies on Vivaldi, a representative distributed NC system, show that under node churn the prediction accuracy of the NC system will be seriously impaired. In this paper, we focus on how to handle the impact of node churn in Vivaldi. Firstly, we propose a simple solution by directly increasing the measurement frequency. Our experiments have demonstrated that this approach can reduce the harm of node churn. However, it increases the communication overhead as the measurement frequency grows. To avoid such expensive solution, we propose the design and implementation of Myth, a decentralised and fast convergence NC system. It introduces the merit of Landmark-based NC system to shorten convergence time in Vivaldi with slight extra overhead. Our experimental results show that Myth outperforms Vivaldi a lot under node churn, without compromising the performance under stable environment. Moreover, we have found that the use of Myth is a cost-effective way to achieve higher prediction accuracy; it will not only improve the prediction accuracy but also save the communication overhead.