Tutorial 1 and Tutorial 2
when?	June 6, 2004 ( Sunday, in the afternoon )
where?	Institute of Computer Science AGH     How to get there... building C-1 (see map)
how much?	20 / 30 Euro ( early / late payment )
Tutorial 3
when?	June 9, 2004 ( Wednesday, 10.50 - 12.30 )
where?	AGH, building C-1, room 406 (IV floor)      (see map)
Tutorial 4
when?	June 9, 2004 ( Wednesday, 14.30 - 16.10 )
where?	AGH, building D-10, room 18 (groundfloor)      (see map)
Tutorial 5
when?	June 8, 2004 ( Tuesday, 16.40 - 18.20 )
where?	AGH, building C-1, room 406 (IV floor)      (see map)

Participants who are interested in attending Tutorial 3 and Tutorial 5 should register prior to the tutorial at the Information Desk. Please note that since tutorials will include hands-on exercises, the number of participants is limited to 35 persons for each tutorial.

Pawel Plaszczak, Krzysztof Wilk, Gridwise Technologies Practical Introduction to Grid, Grid Services Globus Toolkit 3 Abstract: Although Globus Toolkit is the most frequently used Grid middleware, its effective operation normally requires many days of training. For those who cannot afford that much time, Globus remains a mystery. Such people should come to our tutorial. Participants will witness in real time creation of Grid services and will get a feel of what Globus is like, before they decide to try it themselves. During the first hour of the tutorial we will present the prespective of Grid services, and explain the relation between Web services, Grid services, OGSI, OGSA and Globus Toolkit. We will later discuss the Java architecture of Globus 3.0 which implements an OGSI services container framework. The remaining part of the tutorial will consist of a real time programming session, where you will see Java Grid services being implemented, tested and put to work together in an ad-hoc Service-Oriented Architecture. We will demonstrate how to implement several features typical for Grid services, like lifetime management, service data, notifications, and more. Furthermore, we will introduce you to upcoming changes in the Globus Toolkit including the pending release of GT4. In this part of this tutorial we will put emphasis on adopting WS - Resource Framework (WS-RF) and implications of the expected merge of grid services with web services technology. Due to the limited time, this tutorial does not involve hands-on excercises for everyone. You are welcome to bring in your laptop (preferrably Linux, Java 1.3.x or 1.4.x, Globus 3 Core as well as ant, junit, log4j preinstalled) and follow the actions of the instructor, but due to the number of participants we cannot offer personal assistance. Guidelines and code examples will be provided for download. Information will be posted to: http://www.gridwisetech.com/events.

T u t o r i a l   2

Marian Bubak, ICS and ACC Cyfronet AGH Grzegorz Mlynarczyk, WebSoft Software Engineering Methods for Computational Science Abstract: In the last decade a very fast growth of demand for IT systems could have been observed. Growing demand forces an immediate response to users³ needs from the software vendors. The fast growth of demand and needs is connected to a large extent with a significant growth of application complexity with the time needed for implementation getting shorter at the same time. Despite a constant growth of IT market (U.S. $200 milliard) about 50-75% (Hailpern and Santhanam, 2002) of the total software manufacturing costs is destined for quality improvement related to testing, verification and fixing of the occuring errors. It is also estimated that about 30% of software maintenance costs is related exclusively to fixing bugs. Within the tutorial we would like to present methods and devices which improve software production process and particularly tools which reduce costs of software verification as well as an early error detection and localization. We will present briefly to the workshop participants the most important issues as well as attributes related to the quality of software, commonly used evaluation models of software production processes as well as advantages of iterative approach and modern methodologies such as eXtreme Programming or Test-Driven development. In particular, we will focus on issues related to software testing as well as source code static analysis as major components, the proper use of which, in our opinion, enables obtaining a real growth of the quality of provided software. The workshop participants will have an opportunity to get acquainted with the tools which support software production process (mainly OpenSource or freely licensed) at its various stages with a particular attention paid to tools destined just for creating, starting and automation of tests as well as code analysis. Within the workshop we will of course present only a certain part of a total system of software quality assurance. Nevertheless, in our opinion, it is one of the most significant components which create a quality policy. Moreover, an adequate application of the methods and tools in question enables using the described approaches in IT projects regardless of the domain and the method of implementation chosen. We are convinced that the methods may be used successfully both in small and big, spread in many localizations, IT undertakings. Due to a high popularity of realization of IT systems by subcontractors in outsourcing model, we think that this type of methods may also be very widely used in such projects. Because many projects in academic environment are realized by large groups of specialists who have wide, but often, different experience and work standards within the scope of software production, we think that the methods and tools presented will enable control and quality improvements of the applications provided by them, which often become a base point and reference for many other IT solutions, and this is why they should be characterized by high quality. Some of the methods and tools that will be presented during the tutorial have been successfully applied to quality assurance process in CrossGrid project http://www.eu-crossgrid.org. Conclusions drawn from this experience will also shortly be presented. Because of time limit, we do not plan practising exercises by all participants of the workshop. However, we encourage everybody to bring their own laptops and practise the examples while discussed by a lecturer. source codes: http://www.websoft.pl/events/iccs2004/

CrossGrid Project Tutorial Abstract: The CrossGrid tutorial comprises presentations and training exercises prepared in order to familiarize the user with the area of Grid computing, being researched by the CrossGrid project. We are going to present tools aimed at both users and Grid application developers. The exercises cover many subjects, from user-friendly utilities for handling Grid jobs, through interactive monitoring of applications and infrastructure, to data access optimization mechanisms. A description of exercises is available at: http://www.eu-crossgrid.org/tutorial Prerequisites for participants: Basic knowledge of the Linux user environment (shell, editor, etc.) and elementary acquaintance with Grid technologies is necessary to complete the exercises. Participants must be pre-registered for the hands-on sessions, since using the CrossGrid testbed requires appropriate security certificates.

Tom Lehmann, Intel Corporation Building InfiniBand* Based High Performance Compute Clusters Using HPC Middleware Packages Abstract: The addition of high speed interconnects to clustered systems gives the user the ability to execute a wider range of applications efficiently. In the past most of the high speed interconnects were expensive compared to the ubiquitous Ethernet connection and available only from a single vendor. The introduction of the InfiniBand* interconnect changes the environment. The current versions of the 4X InfiniBand interconnect products are rapidly coming down in price and available from an ever widening array of vendors. In addition, the performance of the 4X InfiniBand interconnect is at par with or better than most of the proprietary solutions. Students attending this hands-on workshop will learn how to build Xeon(tm) and Itanium(r) 2 processor based clusters that use an InfiniBand interconnect. The following topics are covered: InfiniBand hardware installation and configuration InfiniBand sub-net manager configuration Integration of InfiniBand drivers into a popular middleware package (NPACI-Rocks) HPLinpack benchmark comparison (gigabit Ethernet vs InfiniBand interconnect) It is expected that participants in the tutorial are familiar with Linux, Linux driver integration and benchmarking with HPLinpack.

Dawid Kurzyniec, Tomasz Wrzosek, Tomasz Ampula, Vaidy Sunderam Emory University Cooperative Resource Sharing with H2O and Harness Abstract: Resource sharing across multiple domains is growing in popularity and adoption. This tutorial will describe alternative, lightweight software infrastructures to aggregate heterogeneous resources over multiple networks. The basic software foundation of this approach is the H2O service-oriented framework for distributed computing. It provides a lightweight, general-purpose, configurable platform for resource sharing. Its usage scenarios may range from peer-to peer file sharing applications, through task-farm applications like SETI@home, to full blown scientific collaborations that involve shared use of distributed computing power. In the basic H2O architectural model, resource providers independently share their raw resources (like CPUs, memory, or storage) via software backplanes called kernels. Authorized entities, which may be clients, third-party resellers, or resource owners themselves, can deploy software components called pluglets into those kernels. Subsequently, pluglets provide specialized services that may be thought of as value-added, virtualized, services over raw resources supplied by the kernel owner. The H2O framework is Java-based: kernels and pluglets are Java objects. As a consequence, H2O-based distributed systems exemplify the write once, run anywhere paradigm. H2O kernels may run on virtually any platform, and every pluglet can be loaded into any H2O kernel, regardless of the underlying operating system. In applications where machine dependent code must be executed (e.g. for high performance), H2O provides a dynamic, semi-automated way to embed native code libraries into pluglet wrappers, transfer them to the desired location, and deploy the requisite code. In this tutorial, we will highlight the architectural and design features of H2O that make it especially suitable for multi-domain resource sharing. Then, provider-side actions to define sharing policies, security and control mechanisms will be discussed. Pluglet development will be described in depth, illustrating simple and complex facilities using both textbook and production examples. Client access to external resources, including aspects of aggregation will then be discussed in the context of Harness II, that constructs the equivalent of a distributed virtual machine over H2O resources. Programming such aggregates using traditional concurrent models, with fault-tolerant extensions will be briefly outlined. The tutorial will conclude with a description of downloading, installation, and deployment steps, operating instructions and graphical tools for management, and demonstrations.