Ricardo Marcelín-Jiménez

VHDLA flexible simulator for distributed algorithms

We present a discrete-event simulation tool to study distributed algorithms and systems whose entities can be modeled using state machines. Our design is focused on the flexibility to simulate different communications environments and concurrent executions.

Locally-constructed trees for ad-hoc routing

Abstract We present a family of self-stabilizing distributed algorithms to built a spanning tree on the underlying communications graph of an ad-hoc wireless network. Next, based on this principle, we show how to construct overlaying structures which are suitable for routing tasks.

Distributed Dynamic Storage in Wireless Networks

This paper assumes a set of identical wireless hosts, each one aware of its location. The network is described by a unit distance graph whose vertices are points on the plane two of which are connected if their distance is at most one. The goal of this paper is to design local distributed solutions that require a constant number of communication rounds, independently of the network size or diameter. This is achieved through a combination of distributed computing and computational complexity tools. Starting with a unit distance graph, the paper shows: 1. How to extract a triangulated planar spanner; 2. Several algorithms are proposed to construct spanning trees of the triangulation. Also, it is described how to construct three spanning trees of the Delaunay triangulation having pairwise empty intersection, with high probability. These algorithms are interesting in their own right, since trees are a popular structure used by many network algorithms; 3. A load balanced distributed storage strategy on top of the trees is presented, that spreads replicas of data stored in the hosts in a way that the difference between the number of replicas stored by any two hosts is small. Each of the algorithms presented is local, and hence so is the final distributed storage solution, obtained by composing all of them. This implies that the solution adapts very quickly, in constant time, to network topology changes. We present a thorough experimental evaluation of each of the algorithms supporting our claims.

Congestion Control for a Fair Packet Delivery in WSN: From a Complex System Perspective

In this work, we propose that packets travelling across a wireless sensor network (WSN) can be seen as the active agents that make up a complex system, just like a bird flock or a fish school, for instance. From this perspective, the tools and models that have been developed to study this kind of systems have been applied. This is in order to create a distributed congestion control based on a set of simple rules programmed at the nodes of the WSN. Our results show that it is possible to adapt the carried traffic to the network capacity, even under stressing conditions. Also, the network performance shows a smooth degradation when the traffic goes beyond a threshold which is settled by the proposed self-organized control. In contrast, without any control, the network collapses before this threshold. The use of the proposed solution provides an effective strategy to address some of the common problems found in WSN deployment by providing a fair packet delivery. In addition, the network congestion is mitigated using adaptive traffic mechanisms based on a satisfaction parameter assessed by each packet which has impact on the global satisfaction of the traffic carried by the WSN.

Cyclic Strategies for Balanced and Fault-Tolerant Distributed Storage

Given a set V of active components in charge of a distributed execution, a storage scheme is a sequence of subsets, B 1,B 2,...,B b , of V where, succesive global states are recorded. The subsets, called blocks, have the same size and are scheduled according to some fixed and cyclic calendar of b steps. During i-th step, block B i is selected. Next, a global snapshot is taken and each component sends its corresponding local state to one of the appointed places in B i , in a way that each component stores (approx.) the same number of local states. Afterwards, if a component of B i crashes, all of the data stored in the block is useless, because the global state can not be reconstructed. In this case, the information recorded in an earlier block can be used to recover a global state, provided there is at least one such block where no component has crashed. The goal is to design storage schema that tolerate as many crashes as possible, while trying to have each component participating in as few blocks as possible and, at the same time, working with large blocks (so that a component in a block stores a small number of local states). In this paper several such schema are described and compared in terms of these measures.

Cluster-Based Localisation Method for Dense WSN: A Distributed Balance between Accuracy and Complexity Fixed by Cluster Size

Localisation is a fundamental requirement for a monitoring and tracking system based on wireless sensor networks (WSN). In order to build an accurate set of measurements, sensor nodes must have information regarding their own position within a system of coordinates. When a considerable number of nodes are randomly scattered over a monitoring area, sensor nodes must be part of a self-organised system which provides a set of local position estimates. Nodes participate under very stringent conditions, for example, limited power supply and reduced computational capabilities. This work presents a GPS-free localisation method consisting of four stages that are executed only once during the network initialisation process. These stages are aimed to increase the overall system lifetime by reducing the signalling overhead commonly involved in distributed localisation procedures. The proposed localisation method turns the initial and complex node deployment to several smaller instances by dividing the network into clusters, which can be solved simultaneously based on local resources only. Simulation results show that this approach produces important savings in the involved overall complexity, which can translate into a trade-off between computational cost and localisation accuracy.

WiSe-Nodes: A Family of Node Prototypes for Wireless Sensor Networks

A Wireless sensor network (WSN) consists of a potentially large number of sensor nodes with wireless-communication capabilities that work collaboratively to achieve a common goal. While every WSN application has specific requirements, there are some common basic services that any WSN must offer in order to support applications. There are available commercial WSN nodes offering such basic services, but they are still very expensive. So, we ask ourselves if it is possible to build a WSN node keeping low costs, while at the same time implementing these basic services: the answer is yes. We built sensor nodes and we call them WiSe-Nodes. In this paper we present the architecture of our WiSe-Nodes as well as the details about their hardware modules. We present also an hybrid MAC-routing protocol called WASP that implements the basic services of a WSN. Finally, as a tool for the WSN developers we introduce our simulator YAWSS that provides a controlled test environment where new WSN protocols can be tested and executed before implementing them in hardware.

Locally-Constructed trees for adhoc routing

We present a family of self-stabilizing distributed algorithms to built a spanning tree on the underlaying communications graph of an adhoc wireless network. Next, based on this principle, we show how to construct two overlaying trees which are suitable for routing tasks.

Improving reliability of distributed storage

A storage scheme is a distributed system that coordinates a set of network-attached components. Using this type of solution it is possible to achieve balance in time and space over the involved components. Although this approach was developed to support efficient global states recording, many storage applications, like web hosting or distributed databases, might profit from it to provide highly available and reliable services. We explore the impact of space and information redundancy in order to improve the integrity requirements of files stored according to this management principles.

A performance study of the Babel file system

The performance of a distributed file system is defined by its hardware components, as well as its operational parameters. Even a slight change on a working condition may induce the major impact, for instance, on service response times. In this paper, we propose a set of experiments on the Babel file system using a client that sends requests for either file storage or retrieval, with two file sizes (512 MB and 1 GB). The aim is to achieve the best working conditions. To measure the performance of the Babel system we took the throughput and the response time when any of two operations (storage or retrieval) was running. The analysis of results showed that, for a given set of operational parameters, there is an optimal file size that gets the best out of the systems performance.