Mengkun Yang

A proactive approach to reconstructing overlay multicast trees

Overlay multicast constructs a multicast delivery tree among end hosts. Unlike traditional IP multicast, the non-leaf nodes in the tree are normal end hosts, which are potentially more susceptible to failures than routers and may leave the multicast group voluntarily. In these cases, all downstream nodes are affected. Thus an important problem in overlay multicast is how to recover from node departures in order to minimize the disruption of service to those affected nodes. In this paper, we propose a proactive approach to restore overlay multicast trees. Rather than letting downstream nodes try to find a new parent after a node departure, each non-leaf node pre-calculates a parent-to-be for each of its children. When this non-leaf node is gone, all its children can find their respective new parents immediately. The salient feature of the approach is that each non-leaf node can compute a rescue plan for its children independently, and in most cases, rescue plans from multiple non-leaf nodes can work together for their children when they fail or leave at the same time. We develop a protocol for nodes to communicate with new parents so that the delivery tree can be quickly restored. Extensive simulations demonstrate that our proactive approach can recover from node departures 5 times faster than reactive methods in some cases, and 2 times faster on average.

A model for replica placement in content distribution networks for multimedia applications

The distribution of multimedia files brings new challenges to the problem of replica placement in content distribution networks (CDN) and invalidates several assumptions underlying the existing solutions. In this paper we formulate a new model for the problem of replica placement to accommodate these new characteristics. We perform a theoretical analysis of the cost of distributing multimedia files over CDNs and find out that, contrary to the intuition, deploying as many replicas as possible is not always a good strategy. We then propose several replica placement algorithms that can determine the optimal number of replicas we should select from a given set of potential sites. By simulation we demonstrate that the performance of clients may degrade if we choose too many sites for replica placement.

A proactive tree recovery mechanism for resilient overlay multicast

Overlay multicast constructs a multicast delivery tree among end hosts. Unlike traditional IP multicast, the non-leaf nodes in the tree are normal end hosts, which are potentially more susceptible to failures than routers and may leave the multicast group voluntarily. In these cases, all downstream nodes are affected. Thus, an important problem for making overlay multicast more dependable is how to recover from node departures in order to minimize the disruption of service to those affected nodes. In this paper, we propose a proactive tree recovery mechanism to make the overlay multicast resilient to these failures and unexpected events. Rather than letting downstream nodes try to find a new parent after a node departure, each non-leaf node precalculates a parent-to-be for each of its children. When this non-leaf node is gone, all its children can find their respective new parents immediately. The salient feature of the approach is that rescue plans for multiple non-leaf nodes can work together for their respective children when they fail or leave at the same time. Extensive simulations demonstrate that our proactive approach can recover from node departures much faster than reactive methods, while the quality of trees restored and the cost of recovery are reasonable

A novel approach to improving search efficiency in unstructured peer-to-peer networks

The decentralized peer-to-peer (P2P) technique has been widely used to implement scalable file sharing systems. It organizes nodes in a system into a structured or unstructured network. The advantages of the unstructured P2P systems are that they have lower maintenance complexity and can better adapt to node heterogeneity as well as network dynamics. However, the search process in unstructured systems is not as efficient as in structured P2P systems because the same search message may go through a node multiple times. To facilitate the complex search and improve the search efficiency, we propose a novel approach of assigning identifications to nodes in an unstructured system. Our method can prevent a node from receiving duplicate search messages and retain the low maintenance overhead for the system. The performance evaluations demonstrate that the proposed approach can improve the search efficiency of unstructured P2P systems while keeping the maintenance overhead at a comparable or even lower level, compared with the traditional unstructured systems.

A cooperative failure detection mechanism for overlay multicast

Overlay multicast is widely accepted as an alternative to IP multicast for implementing group communications due to its easy deployment. One important issue to deal with is the node failures or ungraceful departures from the overlay multicast tree. Fast detection is a key to minimize the disruption of service to the affected nodes participating in the multicast session. In this paper, we propose a cooperative failure detection mechanism that can greatly reduce the failure detection time. We quantify three important measures, i.e., the expected detection time, the probability of false failure detection, and the overhead, and study the fundamental tradeoff among them in failure detection mechanisms. The analysis and simulations show that the proposed cooperative failure detection mechanism can significantly reduce the failure detection time while maintaining the probability of false positive at the same level, at the cost of slightly increased overhead.

Cooperative failure detection in overlay multicast

Node failures and ungraceful departures are important issues to be dealt with in overlay multicast. Fast detection is key to minimizing the disruption of service to the affected nodes participating in the multicast session. In this paper, we propose a cooperative failure detection mechanism that can greatly reduce the failure detection time. A significant contribution of the paper is that we quantify three important measures, i.e., the expected detection time, the probability of false failure detection, and the overhead. This allows us to study the fundamental tradeoff among them in the failure detection mechanisms. The analysis and simulations show that the proposed cooperative failure detection mechanism can significantly reduce the failure detection time while maintaining the probability of false positive at the same level, at the cost of slightly increased overhead.

A segmentation-based fine-grained peer sharing technique for delivering large media files in content distribution networks

Delivering large media files over the Internet is a challenging task because it has some unique features that are different from delivering conventional Web documents. In this paper, we propose a fine-grained peer sharing technique for dealing with the problem in the context of content distribution networks. The key difference of the technique from conventional peer-to-peer systems is that the unit of peer sharing is not a complete media file, but at a finer granularity. By doing so, we improve the flexibility of replica servers for handling client requests. We analyze the storage requirement at replica servers and design a scheduling algorithm to coordinate the delivery process from multiple replica servers to a client. Our simulations show that the fine-grained peer sharing approach can reduce the initial latency of clients and the rejection rate of the system significantly over a simple peer sharing method

A framework for allocating clients to rate-constrained multicast servers

Forming multiple multicast groups is a technique often used to deal with the problem of heterogeneous client capacities by allocating clients with different capacities to different groups. In this paper, we investigate the problem of allocating clients to constrained multicast servers, which, similar to clients, have different capacities. We explore several interesting issues caused by these constraints and propose a framework for generating optimal solutions to the allocation problem. We evaluate the method by simulation and show a substantial performance gain of our algorithm over those considering the server constraints separately.

Intra-Session Fairness in Multicast Communications

Multicast communication achieves scalability by sending data to multiple receivers at the same time. Receivers in a multicast session usually share the fate with each other, even though their processing speed and the capacity of the path they use can be quite different. A conventional multicast session usually consists of a single multicast group and the problem is how to set the group rate so that it is fair to both fast and slow receivers, to some extent. In a replicated multicast service, receivers are divided into groups based on their capacities and a multicast session can consist of multiple multicast groups. The question is how to divide receivers into groups exactly and set appropriate group rates so that it is fair to all the receivers. Most of current work focuses on optimizing the social welfare represented as a sum of some performance measures of receivers [Kar et al., 2002; Stoenescu et al., 2003]. In this paper, we define a new concept called intra-session fairness and give an optimal solution that can achieve fairness among receivers in the same session. The goal is to maximize the minimum fairness value of the receivers. The novelty of the framework is that it is independent of the specific definition of the fairness function on individual receivers. We illustrate a layering method to implement the max-min intra-session fair allocation and demonstrate the significant difference in fairness achieved by the maximal social welfare algorithm and the max-min intra-session fairness algorithm.

A fine-grained peer sharing technique for delivering large media files over the internet

In this paper we propose a fine-grained peer sharing technique for delivering large media files in content distribution networks. The replica servers are divided into groups, and those in the same group cooperate with each other. The key difference of the technique from conventional peer to peer systems is that the unit of peer sharing is not a complete media file, but at a finer granularity, in order to increase the flexibility of replica servers for handling client requests. We design a protocol for peers to exchange the information about available resources with each other, and a scheduling algorithm to coordinate the delivery process from multiple replica servers to a client. Our simulations show that the fine-grained peer sharing approach can reduce the initial latency of clients and the rejection rate of the system significantly over a simple peer sharing method.