Bakhta Meroufel

Dynamic Replication Based on Availability and Popularity in the Presence of Failures

The data grid provides geographically distributed resources for large-scale applications. It generates a large set of data. The replication of this data in several sites of the grid is an effective solution for achieving good performance. In this paper we propose an approach of dynamic replication in a hierarchical grid that takes into account crash failures in the system. The replication decision is taken based on two parameters: the availability and popularity of the data. The administrator requires a minimum rate of availability for each piece of data according to its access history in previous periods, but this availability may increase if the demand is high on this data. We also proposed a strategy to keep the desired availability respected even in case of a failure or rarity (no- popularity) of the data. The simulation results show the effectiveness of our replication strategy in terms of response time, the unavailability of requests, and availability

Lightweight coordinated checkpointing in cloud computing

Fault tolerance is actually an essential issue in cloud computing to face failures and minimize their damages. The checkpointing is a powerful fault tolerance technique that consists of saving the transient state of a computation system on a persistent storage from which the execution state can be restarted in case of failure. The coordinated checkpointing is an efficient checkpointing strategy because it is domino effect-free and it needs only the last stored checkpoint to ensure a consistent state. In this paper we propose a lightweight coordinated checkpointing for cloud computing that minimizes the overload of classical coordinated checkpointing by minimizing the number of the participating virtual machines (VMs) in each checkpointing interval. The experimental results prove that our proposal reduces the overload and improves the system performances.

Availability Management in Data Grid

The data grids are highly distributed environments where nodes are geographically distributed across the globe and shared data are generally very large. The use of replication techniques ensure better availability and easy access to data handled in the grids. In this article, we propose a dynamic replication strategy based on availability and popularity, this replication takes into account failures in the system. The minimum degree of replication is specified by a certain probability of availability and the maximum degree is controlled by the popularity of the data, we introduced also the concept of dynamic primary replica that is used to ensure availability without increasing recovery time. We show in this article that the proposed strategy improves the availability of data according to its popularity and at the same time it improves system performance.

Optimization of checkpointing/recovery strategy in cloud computing with adaptive storage management

Cloud Computing is a type of distributed system that is usually based on the services offered to the user based on SLA contract. In this case, the implementation of a fault‐tolerant system that ensures the reliability and the services continuity becomes a major requirement. In this paper, we propose a fault tolerance strategy based on checkpointing and replication. Our approach uses a smart checkpoint infrastructure for cloud computing tasks. The checkpoints are stored in alternative already paid VMs. This allows resuming a task execution faster and cheaper after a node crash. Since checkpoints are distributed and replicated, our approach increases also the system reliability. The experimental results show the effectiveness of the proposed strategy in term of energy consumption, SLA (System Level Aggregation) violation, and reliability.

Clustering by Swarm Intelligence in the Ad-Hoc Networks

An Ad-hoc wireless is composed mainly of mobile hosts that communicate with each other without fixed infrastructure and no central administration. The main problems associated with these networks are the unpredictable mobility of hosts and a modest flow of communication. In this context, a major problem is partitioning the network in groups called clusters, giving a hierarchical organization. This work presents a deterministic self-stabilizing clustering algorithm for Ad-Hoc networks based on PSO (Particle Swarm Optimization). The proposed approach creates clusters and controls the nodes mobility to ensure more stability for the system. To increase the network life and reduce the answer time of user queries, it proposed also a replication strategy based on the non-similarity degree. The simulations show that the cooperative approaches (clustering, mobility and replication) minimize the energy consumption and increase the QoS of the system.