Scalable and Efficient Discovery of Resources, Applications, and Services in P2P Grids

With millions of computers connected to the Internet, and more coming online each day, how can we ignore such a massive pool of resources? Through distributed computing, namely resource sharing and discovery, we are able to harness all those resources and computing power. Therefore, the aim of this report is to construct an efficient and scalable resource discovery mechanism, capable of searching not only for physical resources (e.g. CPU, Memory, etc.), but also services (e.g. facial recognition, high-resolution rendering, etc.) and applications (e.g. ffmpeg video encoder, programming language compilers, etc.) installed on computers connected to the same Peer-to-Peer Grid network. This will be done in a novel way that consists of combining all resource information into attenuated Bloom filters by following naming rules and using namespaces. The research performed here shows that various systems tackled this problem by focusing on each resource type in isolation. This process consisted of analyzing P2P, Grid, and Cycle Sharing systems that either provide (physical) resource discovery or service discovery. Efficient forms to represent data were also researched in order to minimize storage and transmission costs which impact system efficiency and scalability. This report also presents the performance metrics that will be used to evaluate the system w.r.t. our objectives. Such metrics consist of testing the system by simulating typical usage scenarios, in various network configurations, whilst analyzing the number of messages and hops (among others), along with the effect of fine tuning Bloom filter parameters. Only once the system is implemented and evaluated will we be able to determine if our objectives were met, even though the architecture proposed here was designed with efficiency and scalability in mind