Erich Schikuta

Towards a cost model for distributed and replicated data stores

Large, Petabyte-scale data stores need detailed design considerations about distributing and replicating particular parts of the data store in a cost-effective way. Technical issues need to be analysed and, based on these constraints, an optimisation problem can be formulated. In this paper we provide a novel cost model for building a world-wide distributed Petabyte data store which will be in place starting from 2005 at CERN and its collaborating, world-wide distributed institutes. We elaborate on a framework for assessing potential system costs and influences which are essential for the design of the data store.

The BANG-Clustering System: Grid-Based Data Analysis

For the analysis of large images the clustering of the data set is a common technique to identify correlation characteristics of the underlying value space. In this paper a new approach to hierarchical clustering of very large data sets is presented. The BANG-Clustering system presented in this paper is a novel approach to hierarchical data analysis. It is based on the BANG-Clustering method ([Sch96]) and uses a multidimensional grid data structure to organize the value space surrounding the pattern values. The patterns are grouped into blocks and clustered with respect to the blocks by a topological neighbor search algorithm.

Distributed Trust Management for Validating SLA Choreographies

For business workflow automation in a service-enriched environment such as a grid or a cloud, services scattered across heterogeneous Virtual Organizations (VOs) can be aggregated in a producer-consumer manner, building hierarchical structures of added value. In order to preserve the supply chain, the Service Level Agreements (SLAs) corresponding to the underlying choreography of services should also be incrementally aggregated. This cross-VO hierarchical SLA aggregation requires validation, for which a distributed trust system becomes a prerequisite. Elaborating our previous work on rule-based SLA validation, we propose a hybrid distributed trust model. This new model is based on Public Key Infrastructure (PKI) and reputation-based trust systems. It helps preventing SLA violations by identifying violation-prone services at service selection stage and actively contributes in breach management at the time of penalty enforcement.

Aggregation patterns of service level agreements

IT-based Service Markets require an enabling infrastructure to support Service Value Chains and service choreographies resulting from service composition scenarios. This will result into novel business models where services compose together hierarchically in a producer-consumer manner to form service chains of added value. Service Level Agreements (SLAs) are defined at various levels in such a hierarchy to ensure the expected quality of service for different stakeholders. Automation of service composition directly implies the aggregation of their corresponding SLAs. Aggregation of hierarchical SLAs leads to cross-enterprize business networks such as Virtual Enterprize Organizations (VEO), Extended Enterprizes and Value Networks. During the hierarchical aggregation of SLAs, certain SLA information pertaining to different stakeholders is meant to be restricted and can be only partially revealed to a subset of their business partners. Based on our notions of SLA Choreography and SLA-Views, we formally define an aggregation model to enable the automation of hierarchical aggregation of Service Level Agreements. The aggregation model leads to the discovery of various aggregation patterns in context with service composition and business collaboration.

Aggregating Hierarchical Service Level Agreements in Business Value Networks

Business scenarios such as Business Value Networks and Extended Enterprises pose new challenges for service choreographies across heterogeneous Virtual Organizations. In such scenarios, services compose together hierarchically in a producer-consumer manner to form service supply-chains of added value. Service Level Agreements (SLAs) are defined at various levels in this hierarchy to ensure the expected quality of service for different stakeholders. Automation of service composition directly implies the aggregation of their corresponding SLAs. But so far, the aggregation of SLAs has been treated only as a single layer process which is insufficient to complement the hierarchical aggregation of services. In this paper we elaborate on the requirement of a hierarchical aggregation of SLAs corresponding to service choreographies in Business Value Networks. During the hierarchical aggregation of SLAs, certain SLA information pertaining to different stakeholders is meant to be restricted and can be only partially revealed to a subset of their business partners. We introduce the concept of SLA-Views to protect such privacy concerns. We, then formalize the notion of SLA Choreography and define an aggregation model based on SLA-Views to enable the automation of hierarchical aggregation of Service Level Agreements. The aggregation model has been designed to comply with the WS-Agreement standard.

gSET: trust management and secure accounting for business in the grid

We developed gSET as solution for the unsolved problems in the field of dynamic trust management and secure accounting in commercial virtual organizations. gSET establishes trust and privacy between entities in a grid environment by adapting the concept of secure electronic transactions (SET) used for electronic credit card transfers in eBusiness. Trust is necessary for grid participants in a business environment. It is also necessary to support the dynamic manner of real markets. As distinguished function, in opposite to existing mechanisms as GSI/CAS/VOMS, gSET allows the user to obtain access to a service without disclosing his credentials to the service provider. This minimizes the service providers administrative effort needed for user account management. gSET consists of grid services implemented with WSRF/GT4. gSET is an enabling step to make grids a platform for commercial workflows.

NeuroWeb: an Internet-based neural network simulator

The NeuroWeb project is an Internet-based framework for the simulation of neural networks. It aims for using the Internet as a transparent environment to allow users the exchange of information (neural network objects, neural network paradigms) and the exploit of available computing resources for neural network specific tasks (specifically training of neural networks). NeuroWebs design principles are acceptance, homogeneity, and efficiency.

Building a modular service oriented workflow engine

Current workflow engines, often based on WS-BPEL, are monolithic programs to orchestrate a business process or take part in choreography between partners. They cover aspects like the execution of the workflow description derived from the business process, the communication with external services, and the handling of errors. Both, the communication with services as well as basic error handling are covered by static and internal APIs [S. Modafferi, E. Mussi, and B. Pernici (2006)], and thus provide only simple strictly internal means of interaction with running workflows. Introducing new repair algorithms or self-healing behavior require detailed monitoring and modification of all aspects of a workflow execution thus requiring a modularization of the workflow engine itself. In this paper we will present a modular and service oriented workflow engine. Modularity and service-orientedness are realized by integrating a workflow execution engine with means of logging, repair and possibility to invoke external services through simple Web-based APIs.

Solving Very Large Traveling Salesman Problems by SOM Parallelization on Cluster Architectures

This paper describes how to solve very large Traveling- Salesman Problems heuristically by the parallelization of self-organizing maps on cluster architectures. The used way of parallelizing is a sophisticated Structural Data Parallel approach based on the SPMD model. We distinguish between a non-sophisticated and a sophisticated approach for efficient and simple parallelization of the SOMs.

Multicore and GPU Parallelization of Neural Networks for Face Recognition

Abstract Training of Artificial Neural Networks for large data sets is a time consuming task. Various approaches have been proposed to reduce the efforts, many of them by applying parallelization techniques. In this paper we develop and analyze two novel parallel training approaches for Backpropagation neural networks for face recognition. We focus on two specific paralleliza- tion environments, using on the one hand OpenMP on a conventional multithreaded CPU and CUDA on a GPU. Based on our findings we give guidelines for the efficient parallelization of Backpropagation neural networks on multicore and GPU architectures. Additionally, we present a traversal method finding the best combination of learning rate and momentum term by varying the number of hidden neurons supporting the parallelization efforts.