P2P popularity Series 6: Discussion on various P2P Algorithms

With the rapid development of P2P applications, the discovery technology, as the core issue of P2P applications, not only follows the logic of the technology itself, but also is profoundly influenced by the development trend and Demand Trend of some technologies. -
As mentioned above, the DHT discovery technology is fully built on the basis of a deterministic topology structure, which demonstrates guidance on routes in the network and strong control over node and data management in the network. However, the recognition of deterministic structures limits discovery.AlgorithmImprove efficiency. This paper analyzes the current DHT-based discovery algorithm and finds two important parameter degrees (representing the number of neighbors and the capacity of the route table) and link lengths (average path length of the discovery algorithm) there is a relationship between progressive curves.

The researchers used the degree and diameter parameters in graph theory to study DHT discovery algorithms. They found that these DHT discovery algorithms had a progressive curve relationship between degrees and diameters, as shown in. In N node networks, the figure intuitively shows that when the degree is N, the diameter of the algorithm is O (1). When each node maintains only one neighbor, the diameter of the algorithm is O (n ). This is two extreme cases of the relationship between the degree and the diameter. At the same time, it is impossible to analyze the degree of O (d) and the diameter of O (d) based on graph theory.

From the relationship of the progressive curve, we can see that if we want to get a shorter path length, it will inevitably lead to an increase in the degree of power. However, changes in the actual network connection status will lead to an increase in the complexity of maintaining a large number of neighbors. In addition, researchers have demonstrated that the average path length of O (logn) or even O (logn/logn) cannot meet the needs of network applications with dramatic state changes. The root cause of the constraints of the new discovery algorithm on this compromise is DHT's deterministic understanding of the network topology.

In the unstructured P2P system, the discovery technology has always adopted the flood forwarding method. Compared with the DHT heuristic discovery algorithm, the reliability is poor and the network resources are greatly consumed. The latest research covers two aspects: improving the reliability of the discovery algorithm and finding the shortest path in a random graph. That is, a new understanding of overlapping networks. Among them, the small world features and power laws prove that the topology structure of the actual network is neither a completely random graph recognized by the unstructured system nor a deterministic topology structure adopted by the DHT discovery algorithm.

The meaning of the power law distribution embodied in the actual network can be simply explained as there are a few nodes in the network with a higher degree, while most nodes have a lower degree. A node with a higher degree has more connections with other nodes. It has a higher probability of finding information to be queried.

Features of the small-world [a] [B] model: the network topology has the characteristics of high clustering and short chain. In a network model that complies with the small world feature, nodes can be divided into several clusters based on the Clustering Degree of nodes ), each cluster has at least one node with the highest degree as the center node. A large number of studies have proved that the P2P network represented by Gnutella conforms to the small world feature, that is, there are a large number of highly connected nodes in the network, and some nodes have "Short chains" between them.

Therefore, how to shorten the path length in P2P discovery algorithms becomes how to find these "Short chains. How to generate and find "Short chain" is a new idea of discovering algorithm design, especially in DHT discovery algorithms. The introduction of small world features has a significant impact on P2P discovery algorithms.