Discovery in a Peer-to-Peer Network

General description about Peer-to-Peer discovery

Recently peer-to-peer applications have been widely used for the purposes of sharing resources such as computation and storage. Growth of file sharing applications (e.g., Napster, Gnutella, KaZaA, WinMX, etc) has been remarkably rapid, and numerous system architectures for sharing files have been proposed and implemented in both academia and industry.

Peer-to-peer system architecture produces the need for a method to search for network objects (data, applications, and users) since peer-to-peer systems have the inherent property of objects being distributed in the network. Discovery mechanisms provide a method to locate objects in a network, while remaining decentralized and adaptive to the network environment.

It is required to design discovery mechanisms to meet the primary requirements of efficiency, performance, scalability and robustness.

Efficiency:  The discovery mechanisms are required to be efficient. The peer-to-peer system is expected to accommodate a large number of network objects (e.g., data, applications and users) that have limited capability in their communication, processing, and storage. Discovery mechanisms, thus, need to be efficient so that the operation of discovery mechanisms does not overload network objects and network bandwidth

Performance: The discovery mechanisms need to return meaningful results within an acceptable amount of time.

Scalability: The discovery mechanisms need to perform with reasonable performance when certain parameters of the network increase such as the number of network nodes, the network usage, the number of users, the number of applications, or the amount of data.

Robustness: The discovery mechanisms need to perform with reasonable performance under a network environment in which various types of errors occur (transmission error, loss of network objects, etc).

Current projects

1) Generic Discovery Framework

Numerous discovery mechanisms have already been proposed in many literatures. However, let us say that no existing discovery mechanism is perfect. No existing discovery mechanism can meet all requirements mentioned above at the same time. This fact reminds us the importance of understanding exactly the pros and cons of each discovery mechanism and choosing the mechanism that best fits the given network environment. Therefore, we propose to create a generic discovery framework that provides many options of discovery mechanisms for users, and allows users to select an appropriate discovery mechanism from the options.

Currently we are working on a survey of existing discovery mechanism, making the advantages and disadvantages clear, so that we can classify the existing discovery mechanisms based on appropriate criteria. Afterwards, we are going to use simulations to evaluation some of the discovery mechanisms using appropriate metrics and parameters.

2) P2P discovery in a highly dynamic network environment

Currently we are conducting collaborative research with Denso, a well-known Japanese corporation/laboratory that manufactures systems and products for automobiles. We envision future peer-to-peer application scenarios that involve many mobile users and cars that may join the network and provide or enjoy services in the network. The main issue that we are facing is how to deal with a highly dynamic physical network that includes cell phone network infrastructure, Wireless LAN, or inter-vehicle communication in addition to existing wired networks. The issue is challenging and very significant for the future peer-to-peer applications.

Work and qualification

Project 1 involves intense survey about existing peer-to-peer discovery mechanisms. Students are required to read related papers in a “critical” manner, and understand them well enough to be able to discuss each system in detail. Project 2 requires students to have some creativity and novel ideas in order to attack issues. Therefore, it is a prerequisite for students to know basic network knowledge and preferably have already taken ICS 153 Computer network with a decent grade. Furthermore, students are required to implement simulators/modules in order to evaluate performance of existing discovery mechanisms in both project 1 and 2. Hence, it is strongly recommended for the student to have proficiency in programming, specifically JAVA.  It is also required that a student participates in this project for at least 2 quarters.