Dynamic P2P Systems Research Articles

On Static and Dynamic Partitioning Behavior of Large-Scale P2P Networks

In this paper, we analyze the problem of network disconnection in the context of large-scale P2P networks and understand how both static and dynamic patterns of node failure affect the resilience of such graphs. We start by applying classical results from random graph theory to show that a large variety of deterministic and random P2P graphs almost surely (i.e., with probability 1- <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">o</i> (1)) remain connected under random failure if and only if they have no isolated nodes. This simple, yet powerful, result subsequently allows us to derive in closed-form the probability that a P2P network develops isolated nodes, and therefore partitions, under both types of node failure. We finish the paper by demonstrating that our models match simulations very well and that dynamic P2P systems are extremely resilient under node churn as long as the neighbor replacement delay is much smaller than the average user lifetime.

Resilience of structured P2P systems under churn: The reachable component method

Users in a peer-to-peer (P2P) system join and leave the network in a continuous manner. Understanding the resilience properties of P2P systems under high rate of node churn becomes important. In this work, we first find that a lifetime-based dynamic churn model for a P2P network that has reached stationarity is reducible to a uniform node failure model. This is a simple yet powerful result that bridges the gap between the complex dynamic churn models and the more tractable uniform failure model. We further develop the reachable component method and derive the routing performance of a wide-range of structured P2P systems under varying rates of churn. We find that the de Bruijn graph based routing systems offer excellent resilience under extremely high rate of node turnovers, followed by a group of routing systems that include CAN, Kademlia, Chord and randomized-Chord. We show that our theoretical predictions agree well with large-scale simulation results. We finish by suggesting methods to further improve the routing performance of dynamic P2P systems in the presence of churn and failures.

Load balancing in dynamic structured peer-to-peer systems

Most P2P systems that provide a DHT abstraction distribute objects randomly among “peer nodes” in a way that results in some nodes having Θ(log N) times as many objects as the average node. Further imbalance may result due to nonuniform distribution of objects in the identifier space and a high degree of heterogeneity in object loads and node capacities. Additionally, a node's load may vary greatly over time since the system can experience continuous insertions and deletions of objects, skewed object arrival patterns, and continuous arrival and departure of nodes. In this paper, we propose an algorithm for load balancing in such heterogeneous, dynamic P2P systems. Our simulation results show that in the face of rapid arrivals and departures of objects of widely varying load, our algorithm improves load balance by more than an order of magnitude for system utilizations as high as 80% while incurring an overhead of only about 6%. We also show that our distributed algorithm performs only negligibly worse than a similar centralized algorithm, and that node heterogeneity helps, not hurts, the scalability of our algorithm. Although many of these results are dependent on the workload, we believe the efficiency and performance improvement demonstrated over the case of no load balancing shows that our technique holds promise for deployed systems.