Yves Denneulin

nfsp: a distributed NFS server for clusters of workstations

A consequence of the increasing popularity of Beowulf clusters has been their increasing size (in number of nodes). Yet, the hard drives available on these nodes are only used for the system and temporary files, thus wasting a lot of space (several TiB on large clusters!). The systems that might help recycling this otherwise-unused space are few and far between. This paper presents a NFS server that aims at using the unused disk space spread over the cluster nodes and at offering performance and scalability improvements (compared to the plain NFS servers). The architecture of our solution uses a metaserver and I/O daemons. The client only sees plain NFS, thanks to NFS over UDP spoofing techniques. A first implementation and early performances are shown for this approach.

Performance evaluation of a prototype distributed NFS server

A high-performance file system is normally a key point for large cluster installations, where hundreds or even thousands of nodes frequently need to manage large volumes of data. While most solutions usually make use of dedicated hardware and/or specific distribution and replication protocols, the NFSP (NFS Parallel) project aims at improving performance within a standard NFS client/server system. In this paper we investigate the possibilities of a replication model for the NFS server, which is based on Lasy Release Consistency (LRC). A prototype has been built upon the user-level NFSv2 server and a performance evaluation is carried out.

Distributed semantic caching in grid middleware

This paper proposes a flexible caching solution to improve query evaluation in grids. It reduces both, data transfer and query computation, by adopting a distributed semantic caching approach. Our proposal introduces multi-scale cache cooperation including single site cooperation between object caches and distributed context aware cooperation between several query caches. Different cache miss resolution protocols are introduced for query evaluation and experimented in a grid data management for bioinformatics applications.

INSIGHT: interoperability and service management for the digital home

The emergence of plug-n-play protocols, multimedia and ubiquitous applications is shaping the human habitat into a digital one. The actual plug-n-play device proliferation encourages the development of ubiquitous applications providing the user with a wide set of services to accomplish his everyday tasks. Moreover, the user requires a smooth and transparent interaction with each of his devices and a high quality multimedia experience. However, the digital home is more than ever an heterogeneous and complex environment, devices differ in terms of resources and networking protocols turning the user requirements into an almost impossible task. This paper presents the INSIGHT middleware involving four modules to simplify the digital home complexity. DOXEN provides a cross-device interaction by automatically generating adequate proxies. The DomVision scans the digital home, processes and stores valuable information for management and troubleshooting. The Service Level Checking analyzes the Dom Visions information to provide a personalized service offer for a specific home. And finally, the Resource Manager guarantees the end-user QoS. We also show, through a real experimentation, the ability of INSIGHT to tackle the digital homes complexity.

A synthetic workload generator for cluster computing

Summary form only given. The major issue today on cluster and grid computing is the efficient resource management. The evaluation of scheduling strategies is hard because of the generation of jobs under realistic scenario. This is true for rigid jobs (where the number of processors is fixed) and even more for moldable ones. We present an approach to generate realistic workloads for this kind of jobs. The model we propose is based on the analysis of one year of utilization of the I-cluster, a 225 processors cluster. From this log we extract a typical load for this kind of parallel machines and introduce a way to generate synthetic realistic workloads in an automatic way. This work was done as a way to test scheduling strategies taking into account both rigid and moldable jobs so as the workload generator may handle moldable jobs.

Improving the Performances of a Distributed NFS Implementation

Our NFS implementation, NFSP (NFS Parallele) aims at providing some transparent ways to aggregate unused disk space by means of dividing a usually centralized NFS server into smaller entities: a meta-data server and I/O servers. This paper illustrates the issues related to increasing the performances of such an implementation. Two different approaches have been taken: distributing the load across several servers and implementing the server in a more efficient and intrusive way (in kernel mode). The results obtained with both versions are given and compared to the ones of the first user-mode implementation.

AGIOS: Application-Guided I/O Scheduling for Parallel File Systems

Running data-intensive scientific workflow across multiple data centers faces massive data transfer problem which leads to low efficiency in actual workflow application for scientists. By considering data size and data dependency, we propose a k-means algorithm based initial data placement strategy that places the most related initial data sets into the same data center at workflow preparation stage. During the execution of scientific workflow, by analyzing interdependent relationship between data sets and tasks, we adopt multilevel task replication strategy to reduce volume of intermediate data transfer. The simulation results show that the proposed strategies can effectively reduce data transfer among data centers and improve performance of running data intensive scientific workflows.In remote sensing applications, the traditional procedure for registering a pair of images requires the manual selection of ground control points at significant landmarks of the images. The primary drawback of this approach is that a trained expert is needed to manually select each individual ground control point in the remotely sensed images. This is very laborious and time consuming. In this paper, a new feature-based approach to automate image-to-image registration is proposed for remote sensing applications. This new approach exploits the nonsubsampled contourlet transform to automatically extract a set of control points across spatial and directional resolutions where misalignment between images can be expected to appear. Preliminary experimental results demonstrate the effectiveness and accuracy of the proposed algorithm.For improving the circuit yield, a variety of works based on Dual-Modular Redundancy (DMR) have been studied. Recently, a novel circuit model, Partially Programmable Circuit (PPC), which is made through inserting some Look-Up Tables (LUTs) in conventional logic circuits, has been also proposed. However, they have in common in that they suffer from large area overhead, especially when not all but some faults are specified to be bypassed. Focusing on the fact that an identical copy of parts of the original circuits in DMR and the full programmability of LUTs in PPCs both contain too much functionality for bypassing specified faults, in this paper, we propose three methods to efficiently reduce the functionality of sub-circuits. They can be classified into two approaches, the first one is to reduce the programmability of LUTs in PPCs, and the second one is to apply different ASIC-based sub-circuits, each of which is to bypass a single fault. For the second approach, we propose a novel framework, called Multiple Functional Redundancy (MFR) and its compression method. Experimental results demonstrated that our proposed methods overcome traditional DMR and the original PPC in area overhead while achieving the same yield improvement.

Electromagnetic simulations via parallel computing: an application using scale changing technique for modeling of passive planar reflectarrays in grid environment

Electromagnetic simulation continues to be the primary method by which engineers and researchers analyze and design circuits and systems. Nowadays, the complexification related to the rise of microwaves circuits results in the coexistence of several scales in the same circuit and in the appearance of important dimensions ratios. As the dimensions of modern Electromagnetic circuits continue to shrink, the fabrication and parametric testing of such structures have become more challenging. Increasing market pressures are driving engineers and researchers to minimize the time-to-the-market. So more efficient simulation strategies to accelerate the simulation process are required. Parallel and distributed simulation approaches seem to be a promising approach in this direction. With the growing incidence of computer modeling and simulation, the scope of this paper is to show that parallel simulation of multi-scale models can achieve several orders of magnitude speedup. When mapped onto distributed memory multicomputer systems like Computational Grid, additional speedup is obtained.

Merging file systems and data bases to fit the grid

Grids are widely used by CPU intensive applications requiring to access data with high level queries as well as in a file based manner. Their requirements include accessing data through metadata of different kinds, system or application ones. In addition, grids provide large storage capabilities and support cooperation between sites. However, these solutions are relevant only if they supply good performance. This paper presents Gedeon, a middleware that proposes a hybrid approach for scientific data management for grid infrastructures. This hybrid approach consists in merging distributed files systems and distributed databases functionalities offering thus semantically enriched data management and preserving easiness of use and deployment. Taking advantage of this hybrid approach, advanced cache strategies are deployed at different levels to provide efficiency. Gedeon has been implemented, tested and used in the bioinformatic field.

Distributed file system for clusters and Grids

NFSG aims at providing a solution for file accesses within a cluster of clusters. Criteria of easiness (installation, administration, usage) but also efficiency as well as a minimal hardware and software intrusivity have led our developments. By using several facilities such as distributed file systems (NFSP) and a high-performance data transfer utility (GXfer), we hope to offer a software architecture fully compatible with the ubiquitous NFS protocol. Thanks to a distributed storage (especially multiple I/O servers provided by NFSP), several parallel streams may be used when copying a file from one cluster to another within a same grid. This technique improves data transfers by connecting distributed file system at both ends. The GXfer component implements this functionality. Thus, performances only reachable with dedicated and expensive hardware may be achieved.