A Peer-to-Peer Browsable File Index using a Popularity Based Global Namespace
Thomas Jacobs, Aaron Harwood
Abstract
The distribution of files using decentralized, peer-to-peer (P2P) systems, has significant advantages over centralized approaches. It is however more difficult to settle on the best approach for file sharing. Most file sharing systems are based on query string searches, leading to a relatively simple but inefficient broadcast or to an efficient but relatively complicated index in a structured environment. In this paper we use a browsable peer-to-peer file index consisting of files which serve as directory nodes, interconnecting to form a directory network. We implemented the system based on BitTorrent and Kademlia. The directory network inherits all of the advantages of decentralization and provides browsable, efficient searching. To avoid conflict between users in the P2P system while also imposing no additional restrictions, we allow multiple versions of each directory node to simultaneously exist -- using popularity as the basis for default browsing behavior. Users can freely add files and directory nodes to the network. We show, using a simulation of user behavior and file quality, that the popularity based system consistently leads users to a high quality directory network; above the average quality of user updates. Q
Create a lesson
Related papers
PixelFlow: Token-Level Workload Management for Efficient Distributed DiT Serving
Zhexiang Zhang, Minchen Yu, Yifan Sun et al.
A Kubernetes-Native Request Router for Quality-Aware Inference Serving in the Computing Continuum
Ignjat Karanovic, Pantelis A. Frangoudis, Ivan Čilić et al.
Accelerating Sharded Data Parallelism at Scale with Federated Learning
Gianluca Mittone, Marco Aldinucci
Distributed Edge Inference: an Experimental Study on Multiview Detection
Gianluca Mittone, Giulio Malenza, Marco Aldinucci et al.
P-GADMM: Parallel Group-Based ADMM for Asynchronous Optimization in Heterogeneous Edge Networks
Gaiguo Wei, Qingying Zhang, Heqiang Wang et al.
VERA: Reinforcement Learning for Dynamic Memory Scaling of HPC Workloads in Kubernetes
Ade Pramono, Jie Ren, Ivy Peng