Pawel Wolniewicz

Experiments with Scheduling Divisible Tasks in Clusters of Workstations

We present results of a series of experiments with parallel processing divisible tasks on various cluster of workstations platforms. Divisible task is a new model of scheduling distributed computations. It is assumed that the parallel application can be divided into parts of arbitrary sizes and the parts can be processed independently on distributed computers. Though practical verification of the scheduling model was the primary goal of the experiments also an insight into the behavior and performance of cluster computing platforms has been gained.

Out-of-core divisible load processing

In this paper, we analyze processing divisible loads in systems with a memory hierarchy. Divisible loads are computations that can be divided into parts of arbitrary sizes and these parts can be independently processed in a distributed system. The problem is to partition the load so that the total processing time, including communications and computations, is the shortest possible. Earlier works in the divisible load theory assumed distributed systems with a flat memory model. The dependence of the processing time on the size of the assigned load was assumed to be linear. A new mathematical model relaxing the above two assumptions is proposed in this article. We study distributed systems-which have both the hierarchical memory model and a piecewise linear dependence of the processing time on the size of the assigned load. Performance of such systems is modeled and evaluated. Finally, we compare the efficiency of distributed processing divisible loads in multiinstallment and out-of-core modes. Multiinstallment processing consists in sending multiple small chunks of the load to processors instead of a single chunk which needs external memory. It turns out that multiinstallment is an advantageous strategy for reasonably selected load chunks sizes.

Divisible Load Scheduling in Systems with Limited Memory

In this work we consider scheduling divisible loads on a distributed computing system with limited available memory. The communication delays and heterogeneity of the system are taken into account. The problem studied consists in finding such a distribution of the load that the communication and computation time is the shortest possible. A new robust method is proposed to solve the problem of finding optimal distribution of computations on star network, and networks in which binomial trees can be embedded (meshes, hypercubes, multistage interconnections). We demonstrate that in many cases memory limitations do not restrict efficiency of parallel processing as much as computation and communication speeds.

Optimum divisible load scheduling on heterogeneous stars with limited memory

Scheduling divisible application on a set of heterogeneous processors with limited memory sizes is analyzed in this paper. Divisible loads are computations which allow for dividing computations into several parts of arbitrary sizes, and the parts can be processed independently in parallel. Though the model originated in the parallel computing context, it has strong links with other applications of operations research. A star communication network is assumed. Each processing element of the network is characterized by the processing speed, memory size, speed and startup time of its communication link. The goal is to find a distribution of the load whose schedule length is minimum. The problem is established to be computationally hard. Therefore, two types of algorithms are proposed, and evaluated: an exact algorithm whose execution time may be exponential, and polynomial-time heuristics.

First Prototype of the CrossGrid Testbed

The CrossGrid project is developing new grid middleware components, tools and applications with a special focus on parallel and interactive computing. In order to support the development effort and provide a test infrastructure, an international grid testbed has been deployed across 9 countries. Through the deployment of the testbed and its supporting services, CrossGrid is also contributing to another important project objective, the expansion of the grid coverage in Europe. This paper describes the status of the CrossGrid testbed.

g-Eclipse - an integrated framework to access and maintain Grid resources

The g-Eclipse framework provides a general, integrated workbench toolset for grid users, operators and developers. Based on the open source eclipse ecosystem, g-Eclipse supports scientists to interact with grid resources independent of the underlying grid middleware. Its main objective is to deliver an extensible framework for different grid actors, by providing a unified abstraction of the grid. The grid abstraction enables grid application users to access the grid in a desktop-like manner with wizards specific for common use cases; it also provides a set of visual configuration tools to maintain and configure grid resources.

User Management for Virtual Organizations

Scalable and fine-grained Grid authorization requires moving away from a gridmapfile based access control and 1-to-l mappings to individual OS user accounts. This is recognized and addressed to by virtual organization (VO) authorization services, e. g. VOMS/LCAS and CAS. They, however, do not address user OS account management and isolation/sandboxing requirements, such as flexible pooling of accounts while maintaining auditing records. This paper describes some existing systems for user management for VOs and provides a list of requirements for a new user management system on which our current research is focused on.

International grid CA interworking, peer review and policy management through the european datagrid certification authority coordination group

The Certification Authority Coordination Group in the European DataGrid project has created a large-scale Public Key Infrastructure and the policies and procedures to operate it successfully. The infrastructure demonstrates interoperability of multiple certification authorities (CAs) in a novel system of peer-assessment of the roots of trust. Crucial to the assessment is the definition of minimum requirements that all CAs must meet in order to be accepted. The evaluation is aided by software-generated trust matrices. Related work building on this infrastructure is described. The groups policies and experience now form the basis of the new European Policy Management Authority for Grid Authentication in e-Science.

g-Eclipse --- A Contextualised Framework for Grid Users, Grid Resource Providers and Grid Application Developers

As the future pervasive and ubiquitous computing environment will be composed of resources from local computing, Grid, SOA and Web infrastructures, the complexity of this distributed system will increase significantly. At the same time the end user wants easy and simple access to his computing environment while he receives more responsibilities i.e. for the composition and the management of the system. In order to perform his daily work, the end user needs a general workbench toolset which supports customisation and contextualisation for the user. The g-Eclipse framework offers an eco-system to access Grid infrastructures with support for contextualised user roles. Currently, the g-Eclipse framework includes contextualised perspectives for Grid end users, Grid resource provider and Grid application developers. The abstraction layer of the g-Eclipse system, its integration in the Eclipse framework and the main use cases are presented.

Experience with the international testbed in the crossgrid project

The International Testbed of the CrossGrid Project has been in operation for the last three years, including 16 sites in 9 countries across Europe. The main achievements in installation and operation are described, and also the substantial experience gained on providing support to application and middleware developers in the project. Results are presented showing the availability of a realistic Grid framework to execute distributed interactive and parallel jobs.