A file system for wireless mesh networks

Over the last years, research has focused on Wireless Mesh Networks (WMNs) to bridge the gap between cellular networks, Mobile Ad-hoc Networks (MANETs) and Wireless Local Area Networks (WLANs). WMNs use WLAN technology to create a multi-hop ad-hoc network, which uses packet forwarding to allow nodes of the same network to communicate with each other, even though they are not within direct communication range. They also introduce static devices to the MANET infrastructure to create a backbone network of fixed WLAN base stations. As a result, WMNs combine the advantages of WLAN networks with those of cellular networks, i. e. e. g. low set-up costs, seamless access, and broad network coverage. However, WMN also require new communication protocols and new approaches for distributed applications. In my thesis, I analyze the architecture and the features of six different distributed file systems as a solution for users to collaborate in a WMN. The evaluation shows that the six file systems lack important features. Therefore, I describe in this thesis a new approach for a distributed file system for WMNs—the Wireless File-System (WFS). To this end, my work first focuses on the mechanisms of Peer-to-Peer (P2P) based file systems, as frequent topology changes and the fluctuation of network participants in mesh networks require a flexible solution. By design, P2P systems are self-organizing, distribute the system workload and provide services to other nodes, but also require new approaches to guarantee the confidentiality and the availability of the shared data. However, my thesis shows that P2P based file systems do not work in WMNs. It also identifies which requirements exist for a distributed file system in WMNs. First, I have to minimize the communication overhead for every file system operation while still guaranteeing the availability of critical system data. Second, I have to secure all user data. To solve these problems, I consider the communication overhead and the security for every structure of the file system. My approach is based on my new architecture for a distributed file system. One important aspect of this architecture is the introduction of three different data types: First, containers that store the metadata and the security information for the file system, every directory and every file. Second, cell trees that combine Merkle-trees and B+-trees to provide a dynamic data structure that stores the content of the different file system items. To minimize dependencies during updates, each file system item uses a unique cell tree. Third, the inode tree, which stores a link to the latest version of each file system item. A critical aspect of the file system is to secure cell trees against unauthorized access or manipulation and to still allow users to share access to containers and cell trees with other trusted users. At the same time, the inode tree stores important information about the status of the file system and must be readable by everyone while still being resilient against manipulation. My solution to this problem is a new cryptographically enforced access control system, which includes a user management system and a key distribution scheme.