Catálogo de publicaciones - libros

Utility computing is a service provisioning model which will provide adaptive, flexible and simple access to computing resources, enabling a pay-per-use model for computing similar to traditional utilities such as water, gas or electricity. On the other hand, grid technology provides standard functionality for flexible integration of diverse distributed resources. This paper describes and evaluates an innovative solution for utility computing, based on grid federation, which can be easily deployed on any infrastructure based on the Globus Toolkit. This solution exhibits many advantages in terms of security, scalability and site autonomy, and achieves good performance, as shown by results, mainly with compute-intensive applications.

Even though with few exceptions, grid workloads are dominated by single-node jobs, not all of these jobs are necessarily independent or unrelated. For instance, sets of jobs may be grouped because they are submitted by users in batches, e.g., to perform parameter sweeps. However, there is no reported data to confirm the presence and structure of these groupings, despite the large potential impact of such information. To address this lack of information, in this work we present a first investigation into the characteristics of groups of jobs present in grid workloads. First, we define three types of job groupings: batch, continued, and bursty submissions. Then, we analyze the characteristics of these groupings for three long-term traces from currently deployed grid environments. Notably, our results show that the various groupings are responsible for up to 96% of the total CPU time consumption. Finally, we present insights into the performance of real grids in dealing with grouped jobs.

Using grid resources to execute scientific applications requiring a large amount of computing power is attractive but not easy from the user’s point of view. Vigne is a Grid system designed to provide users with a simplified view of a grid. This paper presents a set of system services that allow to run a wide range of distributed applications in a simple and efficient manner. A running prototype has been implemented as a proof of concept and experiments on the Grid’5000 testbed show the efficiency of our approach.

The interest among a geographically distributed user base to mine massive collections of scientific data propels the need for efficient data dissemination solutions. An optimal data distribution scheme will find the delicate and often application-specific balance among conflicting success metrics such as minimizing transfer times, minimizing the impact on the network, and uniformly distributing load among participants. We use simulations to explore the performance of classes of data-distribution techniques, some of which successfully deployed in large peer-to-peer communities, in the context of today’s data-centric scientific collaborations. Based on these simulations we derive several recommendations for data distribution in real-world science collaborations.

Workflow applications executed in Grid environments are not able to take advantage of all the potential parallelism they might have. This limitation in the usage of parallelism comes from the fact that when there is a producer/consumer situation communicating using files, the consumer does not start its execution till the producer has finished creating the file to be consumed, and the file has been copied to the consumer (if needed).

In this paper, we propose a publish/subscribe mechanism that allows consumers to read the file at the same time it is being produced. In addition, this mechanism is implemented in a transparent way to the application, so does not require any special feature from the local filesystems.

Finally, we show that our mechanisms can speedup applications significantly. In our best test we divided by two the execution time of some applications, but other applications may have even higher benefits.

Grid applications need to be fault tolerant, malleable, and migratable. In previous work, we have presented , an efficient mechanism addressing these issues for divide-and-conquer applications. In this paper, we present a mechanism for writing partial results to checkpoint files, adding the capability to also tolerate the of all processors, and to allow suspending and later resuming an application.

Both mechanisms have only negligible overheads in the absence of faults, even with extremely short checkpointing intervals like one minute. In the case of faults, the new checkpointing mechanism outperforms orphan saving by 10% to 15%. Also, suspending/resuming an application has only little overhead, making our approach very attractive for writing grid applications.

The grid environment presents numerous opportunities for business applications as well as for scientific ones. Nevertheless the current trends seem to lead to several independent specialized grids in opposition to the early visions of one generic world wide grid. In such a cross-grid context, the environment might be harder to manipulate whereas more decisions must be handled from user-side. Our proposal is a distance-based decision-making support designed to be usable, adaptable and accurate. Our main contribution is to ensure the profitability of classical monitoring solutions by improving their usability. Our approach is illustrated and validated with experiments in a real grid environment.

In Grid systems, nontrivial qualities of service have to be provided to users by the resource providers. However, resource management in a decentralized infrastructure is a complex task as it has to cope with different policies and objectives of the different parties: providers and consumers/users. Agreement-based resource management is considered to solve many of these problems as the conflicts between the users and resource providers can be reconciled in a negotiation process. Such negotiation processes must be automated with no or minimal human interaction, considering the potential scale of Grid systems and the amount of necessary transactions. Therefore, strategic negotiation models and strategic negotiation strategies play important roles. In this paper, negotiation strategies considering time and opportunity functions for Grid scheduling are proposed and examined. The simulation results demonstrate that the negotiation strategies are suitable and effective for Grid environments.

This paper introduces the design of , a new Grid framework that will, for the first time, allow multiple invocations of symbolic computing applications to interact via the Grid. is designed to support the specific needs of symbolic computation, including computational steering (greater interactivity), complex data structures, and domain-specific computational patterns (for irregular parallelism). A key issue is heterogeneity: is designed to orchestrate components from different symbolic systems into a single coherent (possibly parallel) Grid application, building on the standard for data exchange between mathematically-oriented applications. The work is being developed as part of a major EU infrastructure project.

This paper outlines a feasible approach to ranking Grid resources based on an easily obtainable application-specific performance model utilizing low-level performance metrics. First, Grid resources are characterized using low-level performance metrics; Then the performance of a given application is associated to the low-level performance measurements via a ; Finally, the Ranking Function is used to rank all available resources on the Grid with respect to the specific application at hand. We show that this approach yields accurate results.