Catálogo de publicaciones - libros

The Broker with the cost function model of the ISS/VIOLA Meta-Scheduling System implementation is described in details. The Broker includes all the algorithmic steps needed to determine a well suited machine for an application component. This judicious choice is based on a deterministic cost function model including a set of parameters that can be adapted to policies set up by computing centres or application owners. All the quantities needed for the cost function can be found in the DataWarehouse, or are available through the schedulers of the different machines forming the Grid. An ISS-Simulator has been designed to simulate the real-life scheduling of existent clusters and to virtually include new parallel machines. It will be used to validate the cost model and to tune the different free parameters.

In order to foster uptake by scientific and business users we need an easy way to access Grid resources. This is the motivation for the A-WARE project. We build upon a fabric layer of Grid and other resources, by providing a higher-layer service for managing the interaction with these resources - A One-Stop, Fire-and-(almost)Forget, Dropping-off and Rendezvous Point. Work assignments can be formulated using domain specific dialects, allowing users to express themselves in their domain of expertise. Both Web service and REST bindings are provided, as well as allowing the component to be embedded into other presentation technologies (such as portals). In addition common desktop notification mechanisms such as Email, RSS/Atom feeds and instant messaging keep users informed and in control. We propose using the Java Business Integration specification as the framework for building such a higher-level component, delivering unprecedented opportunities for the integration of Grid technologies with the enterprise computing infrastructures commonly found in businesses.

Our energy production increasingly depends on regenerative energy sources, which impose new challenges. One problem is the availability of regenerative energy sources like wind and solar radiation that is influenced by fluctuating meteorological conditions. Thus the development of forecast methods capable of determining the level of power generation (e.g., through wind or solar power) in near real-time is needed. Another scenario is the determination of optimal locations for power plants. These aspects are considered in the domain of energy meteorology. For that purpose large data repositories from many heterogeneous sources (e.g., satellites, earth stations, and data archives) are the base for complex computations. The idea is to parallelize these computations in order to obtain significant speed-ups. This paper reports on employing Grid technologies within an ongoing project, which aims to set up a Grid infrastructure among several geographically distributed project partners. An approach to transfer large data sets from many heterogenous data sources and a means of utilizing parallelization are presented. For this purpose we are evaluating various Grid middleware platforms. In this paper we report on our experience with Globus Toolkit 4, Condor, and our first experiments with UNICORE.

Sequence analysis is one of the most fundamental tasks in molecular biology. Because of the increasing number of sequences we still need more computing power. One of the solutions are grid environments, which make use of computing centers. In this paper we present plug-in which enables the use of BLAST software for sequence analysis within Grid environments such as UNICORE (Uniform Interface to Computing Resources) and GPE (Grid Programming Environment).

This paper describes the development of a Job Management Enterprise Application (JMEA) which was developed by the DEISA material science and plasma physics joint research activities. It is capable of submitting jobs to a UNICORE server infrastructure and managing them. Since it is a Java EE application, it can be used by multiple users concurrently. Furthermore, it prefetches and caches request results in order to able of responding as quick as possible to client requests. In addition to normal user credentials it also supports the use of proxy credentials and explicit trust delegation.

DEISA is a consortium of leading national supercomputing centers that is building and operating a persistent distributed supercomputing environment with continental scope in Europe. To integrate their resources, the DEISA partners have adopted the most advanced middleware and applications currently available. The consortium decided to embrace UNICORE as a job submission interface for the DEISA grid infrastructure. UNICORE is the foremost European grid technology able to hide the complexity of the underlying resources providing a user-friendly graphical user interface for job submission. This paper presents the deployment solution and strategies implemented by DEISA in order to adapt UNICORE for their infrastructure.

Multiple plans to create petascale computing environments have been announced. This workshop addressed what bioinformatics or computational biology applications can or should accomplish with such facilities, and what obstacles must be overcome in order to implement and use effective and important problems in the life sciences (biology, biochemistry, environmental sciences, etc.).

This paper describes some of the issues involved with scaling biomolecular simulations onto massively parallel machines drawing on the Blue Matter application team’s experiences with Blue Gene/L. Our experiences in scaling biomolecular simulation to one atom/node on BG/L should be relevant to scaling biomolecular simulations onto larger peta-scale platforms because the path to increased performance is through the exploitation of increased concurrency so that even larger systems will have to operate in the extreme strong scaling regime. Petascale platforms also present challenges with regard to the correctness of biomolecular simulations since longer time-scale simulations are more likely to encounter significant energy drift. Total energy drift data for a microsecond-scale simulation is presented along with the measured scalability of various components of a molecular dynamics time-step.

Petascale computing is currently a common topic of discussion in the high performance computing community. Biological applications, particularly protein folding, are often given as examples of the need for petascale computing. There are at present biological applications that scale to execution rates of approximately 55 teraflops on a special-purpose supercomputer and 2.2 teraflops on a general-purpose supercomputer. In comparison, Qbox, a molecular dynamics code used to model metals, has an achieved performance of 207.3 teraflops. It may be useful to increase the extent to which operation rates and total calculations are reported in discussion of biological applications, and use total operations (integer and floating point combined) rather than (or in addition to) floating point operations as the unit of measure. Increased reporting of such metrics will enable better tracking of progress as the research community strives for the insights that will be enabled by petascale computing.

An important issue in computational biology is the reverse engineering problem for genetic networks. In this ongoing work we consider reverse engineering in the context of univariate finite fields models. A solution to the reverse engineering problem using multipoint interpolation relies on intensive arithmetic computations over finite fields, where multiplication is the dominant operation. In this work, we develop an efficient multiplier for fields GF (2^ m ) generated by irreducible trinomials of the form α ^ m  +  α ^ n  + 1. We propose a design described by a parallel/serial architecture that computes a multiplication in m clock cycles. This approach exploits symmetries in Mastrovito matrices in order to improve time complexities of an FPGA (Field Programmable Gate Array) implementation. According to preliminary performance results, our approach performs efficiently for large fields and has potential for an efficient solution of the reverse engineering problem for large genetic networks, as well as other finite fields applications such as cryptography and Reed-Solomon decoders.