Boris Mejias

A Relaxed-Ring for Self-Organising and Fault-Tolerant Peer-to-Peer Networks

There is no doubt about the increase in popularity of decentralised systems over the classical client-server architecture in distributed applications. These systems are developed mainly as peer-to-peer networks where it is possible to observe many strategies to organise the peers. The most popular one for structured networks is the ring topology. Despite many advantages offered by this topology, the maintenance of the ring is very costly, being difficult to guarantee lookup consistency and fault-tolerance all the time. By increasing self-management in the system we are able to deal with these issues. We model ring maintenance as a self- organising and self-healing system using feedback loops. As a result, we introduce a novel relaxed-ring topology that is able to provide fault-tolerance with realistic assumptions concerning failure detection. Limitations related to failure handling are clearly identified, providing strong guarantees to develop applications on top of the relaxed-ring architecture. Besides permanent failures, the paper analyses temporary failures and broken links, which are often ignored.

Beernet: Building Self-Managing Decentralized Systems with Replicated Transactional Storage

Distributed systems with a centralized architecture present the well known problems of single point of failure and single point of congestion; therefore, they do not scale. Decentralized systems, especially as peer-to-peer networks, are gaining popularity because they scale well, and do not need a server to work. However, their complexity is higher due to the lack of a single point of control and synchronization, and because consistent decentralized storage is difficult to maintain when data constantly evolves. Self-management is a way of handling this higher complexity. In this paper, the authors present a decentralized system built with a structured overlay network that is self-organized and self-healing, providing a transactional replicated storage for small or large scale systems.

THE RELAXED-RING: A FAULT-TOLERANT TOPOLOGY FOR STRUCTURED OVERLAY NETWORKS

Fault-tolerance and lookup consistency are considered crucial properties for building applications on top of structured overlay networks. Many of these networks use the ring topology for the organization or their peers. The network must handle multiple joins, leaves and failures of peers while keeping the connection between every pair of successor-predecessor correct. This property makes the maintenance of the ring very costly and temporarily impossible to achieve, requiring periodic stabilization for fixing the ring. We introduce the relaxed-ring topology that does not rely on a perfect successor-predecessor relationship and it does not need a any periodic maintenance. Leaves and failures are considered as the same type of event providing a fault-tolerant and self-organizing maintenance of the ring. Relaxed-rings limitations with respect to failure handling are formally identified, providing strong guarantees to develop applications on top of the architecture. Besides permanent failures, the paper analyses temporary failures and false suspicions caused by broken links, which are often ignored.

WSN and P2P: A Self-Managing Marriage

Wireless sensor networks are designed for a very wide, yet specific, purpose. Their components have processing and power limitations. Due to these limitations, decisions by running complex algorithms with the information collected by sensors must be done in components external to the WSN. This document presents a combination of WSN and p2p networks to ease the development of systems that rely on WSN functionality. As a result, we propose the creation of a programming abstraction that allows developers to concentrate on the functionality of the developing system. We also propose the use of feedback loops as a way to design and develop the components of the abstraction and to define self-managing behavior for them. Those components should be also lowly coupled, interchangeable and extensible.

From Mini-clouds to Cloud Computing

Cloud computing has many definitions with different views within industry and academia, but everybody agrees on that cloud computing is the way of making possible the dream of unlimited computing power with high availability. However, being a cloud computing provider seems to be reserved to very large companies that can achieve having a huge data center. The rest of the companies and institutions have to play the role of cloud users. We propose an architecture to organize a set of mini-clouds provided by different institutions, in order to provide a larger cloud that appears to its users as a single one. Such architecture requires selfmanaging behaviour in order to deal with the complexity of matching cloud users requests with the computing utility that mini-clouds of institutions can offer.

Visualizing Transactional Algorithms for DHTs

The focus of this demonstrator is on the study of algorithms for implementing transactions on peer-to-peer networks. Their visualization contributes to the analysis and test of the protocols, verifying their tolerance to failures. In particular, we show a DHT running two-phase commit and the Paxos consensus algorithm.

Improving the Peer-to-Peer Ring for Building Fault-Tolerant Grids

Peer-to-peer networks are gaining popularity in order to build Grid systems. Among different approaches, structured overlay networks using ring topology are the most preferred ones. However, one of the main problems of peer-to-peer rings is to guarantee lookup consistency in presence of multiple joins, leaves and failures nodes. Since lookup consistency and fault-tolerance are crucial properties for building Grids or any application, these issues cannot be avoided. We introduce a novel relaxed-ring architecture for fault-tolerant and cost-efficient ring maintenance. Limitations related to failure handling are formally identified, providing strong guarantees to develop applications on top of the relaxed-ring architecture. Besides permanent failures, the paper analyses temporary failures and broken links, which are often ignored.

PEPINO: PEer-to-Peer network INspectOr

PEPINO is a simple and effective peer-to-peer network inspector. It visualises not only meaningful pointers and connections between peers, but also the exchange of messages between them, providing a useful tool for debugging purposes. It can monitor running networks, simulate them and log them in order to reproduce interesting case scenarios. Failures can be explicitly introduced to study fault tolerant algorithms. The graphical representation of the network uses a physical model to attract or repel peers, allowing the user to study the system from different points of view. This demo aims to present the use of PEPINO in the development of a novel relaxed-ring topology for fault tolerant networks, where the representation of the ring based on predecessors may differ from the ring based on successors. We show how PEPINO is also useful for visualising other network topologies such as perfect ring or unstructured networks.

A software based approach to eliminate all SEU effects from mission critical programs

On space missions, software suffers from Single Event Upsets (SEUs), producing bit-flips. A single bitflip can impede the whole program. In this work we present a novel software based approach to tolerate SEUs by redundant execution and protected memory. This approach can protect specifically designed programs against all SEUs. Our aim is to be able to use of-the-shelf processors and other complex circuits by designing appropriately the software and simple hardened devices (e.g. memory). A Linux based simulator has been developed and implemented: it uses, validates and allowed us to refine our hypotheses. A hardware prototype confirmed these results.

Lock-Free Decentralized Storage for Transactional Upgrade Rollback

Installing and upgrading software may introduce conflicts and errors into a system. Transactional Rollback allows the system to return back to a previous, stable and known state. However, to perform such a rollback, it is necessary to store a large amount of information including configuration and installation logs, as well as different versions of software packages. Nevertheless, much of this information is common to several users using the same software and performing the same operations. We can reduce the total amount of storage by having a decentralized architecture using a Distributed Hash Table (DHT) to localise shared resources. We propose a lock-free key/value-set protocol to add and remove data from the DHT. The lock-free protocol is not limited to transactional rollback, and it can be used by other applications that also need value-sets as part of their stored data.