S ECOND PROTOTYPE I N C L U D I N G M E C H A N I S M S U S E D F O R D E F I N I T I O N O F T A S K F L O W, D E V E L O P E D V I S U A L I Z A T I O N T O O L S A N D G R I D C O M E N H A N C E M E N T S Document Filename: Activity: Partner(s): Lead Partner: Document classification: JRA PSNC, VGTU, VU PSNC PUBLIC : Abstract: This document presents the second prototype of components developed within the Joint Research Activity Enhanced Application Services on Sustainable e-infrastructure and deployed on the testbed. For each individual component, a brief description is included, with its main features, technical information about the software and future extensions or improvements. PUBLIC Page 1 of 29
Released for moderation to Approved for delivery by Document review and moderation Name Partner Date Signature Document Log Version Date Summary of changes Author 0.1 7/10/2009 Document template Bartek Palak 0.2 15/10/2009 Visualization tools description Arnas Kačeniauskas 0.3 16/10/2009 Description of Gridcom features Danila Piatov 0.4 19/10/2009 Migrating Desktop release functionality Bartek Palak 1.0 30/10/2009 Review and update Aake Edlund, Mario Kadastik PUBLIC Page 2 of 29
Contents LIST OF ACRONYMS... 5 1. EXECUTIVE SUMMARY... 6 REFERENCES... 7 2. MIGRATING DESKTOP PLATFORM... 8 2.1. OVERVIEW... 8 2.2. PRODUCT DESCRIPTION... 8 2.2.1. Source code... 9 2.2.2. Binary release... 9 2.2.3. Documentation and manuals... 10 2.3. DESCRIPTION OF WORK... 10 2.3.1. Handling glite parametric jobs in Migrating Desktop... 10 2.3.2. Adaptation of mechanisms used for definition and management of task flow... 11 2.3.3. Improvements of advanced data management in Migrating Desktop... 11 2.3.4. Integration of Migrating Desktop and VizLitGrid... 11 2.3.5. Training materials... 11 2.4. PLANS FOR THE NEXT RELEASE... 11 3. GRIDCOM... 12 3.1. OVERVIEW... 12 3.2. PRODUCT DESCRIPTION... 12 3.2.1. Source code... 13 3.2.2. Binary release... 13 3.2.3. Documentation and manuals... 13 3.3. DESCRIPTION OF WORK... 14 3.3.1. User access control... 14 3.3.2. User management interface... 14 3.3.3. Compatibility with BalticGrid certificate policy... 14 3.3.4. System load optimization... 14 3.3.5. Gridcom official web site... 14 3.3.6. Documentation... 14 3.4. PLANS FOR THE NEXT RELEASE... 15 4. PARAVIEW... 16 4.1. OVERVIEW... 16 4.2. PRODUCT DESCRIPTION... 16 4.2.1. Source code... 17 4.2.2. Binary release... 18 4.2.3. Documentation and manuals... 18 4.3. DESCRIPTION OF WORK... 19 5. GRID VISUALIZATION E-SERVICE VIZLITG... 20 5.1. OVERVIEW... 20 5.2. PRODUCT DESCRIPTION... 20 5.2.1. Source code... 21 5.2.2. Binary release... 22 5.2.3. Documentation and manuals... 22 5.3. DESCRIPTION OF WORK... 22 5.4. PLANS FOR THE NEXT RELEASE... 23 6. SPECIALIZED VISUALIZATION TOOL FOR PARTICLE SYSTEMS VISPARTDEM... 24 PUBLIC Page 3 of 29
6.1. OVERVIEW... 24 6.2. PRODUCT DESCRIPTION... 24 6.2.1. Source code... 25 6.2.2. Binary release... 27 6.2.3. Documentation and manuals... 27 6.3. DESCRIPTION OF WORK... 27 7. SUMMARY AND CONCLUSIONS... 29 PUBLIC Page 4 of 29
LIST OF ACRONYMS CE CPU GUI HDF5 JDL MPI NGI PC OS OSGi SDL SE SGM UI UIG VTK WN Computing Element Central Processing Unit Graphical User Interface Hierarchical Data Format Job Description Language Message Passing Interface National Grid Initiative Personal Computer Operating System Open Services Gateway initiativ Simple Directmedia Layer library Storage Element Software Grid Manager User Interface User Interface for Graphics Visualization ToolKit Worker Node PUBLIC Page 5 of 29
1. EXECUTIVE SUMMARY The primary goal of the Joint Research Activity Enhanced Application Services on Sustainable e-infrastructure is an extension of the project infrastructure with advanced services through development of: Migrating Desktop Platform (intuitive interface to project resources), Gridcom (webbased groupware) and special tools for visualization of application in distributed grid environment. Development of main activity components results in preparation of the second release: Second prototype including includes mechanisms used for definition of task flow, developed visualization tools and Gridcom enhancements (DJRA1.4). Migrating Desktop The aim of the Migrating Desktop Platform is to provide grid users and/or application developers with a framework which hides the details of most Grid services. This advanced graphical user interface allows the user to access the Grid resources, run applications, monitor, visualize and manage data files. Works over the second prototype was focused on advanced job types support. Changes in job definition, submission and tracking mechanisms were made to handle glite parametric jobs. Implementation of workflow mechanism from scratch significantly exceeds JRA activity resources, so after detailed evaluation of existing tools, the Kepler workflow orchestration was adopted to support applications based on flow of results. Gridcom Gridcom is a simple Web-based grid interface which supports group access to a workspace and multijob applications. The features prepared for the second prototype include user access control, user management interface, system load optimization, Gridcom site and the documentation. Visualization tools Considering the results of performance analysis (DJRA1.1) development of three different visualization tools has been performed. New versions of developed tools have been deployed on BalticGrid-II testbed for visualization: ParaView 3.6.1 is the most powerful open-source software for distributed parallel visualization of large datasets. ParaView employing GPU rendering has been deployed in this prototype. Parallel reader has been developed and applied for independent reading of decomposed HDF5 files. Visualization e-service VizLitG 0.3 is designed for convenient access and interactive visualization of remote data files located in grid. Deployed release includes revisited security and enhanced interactivity including VTK widgets. Visualization tool for particle systems VisPartDEM 0.1 has been developed and deployed. Advanced functionality for visualization of propagating cracks has been implemented in the VisPartDEM software. PUBLIC Page 6 of 29
REFERENCES [1] Squillacote, A. ParaView Guide. Version 3. Kitware, Inc., (2008). [2] VTK User's Guide Version 5, Inc. Kitware. 5th Edition, (2006). [3] Qt: http://trolltech.com/products/qt/ [4] Python: http://www.python.org/ [5] HDF5: http://hdf.ncsa.uiuc.edu/products/hdf5/ [6] DJRA1.1: Report on Performance analysis of sequential and parallel visualization in grid environment, August 2008. [7] BalticGrid-II: http://www.balticgrid.eu/ [8] GlassFish: https://glassfish.dev.java.net/ [9] GVid: http://www.gup.jku.at/gvid/ [10] glite: http://glite.web.cern.ch/glite/ [11] Marinilli, M. Java Deployment with Jnlp and WebStart. Sams Publishing, (2001). [12] International Euro-Par Conference, Lecture Notes in Computer Science, vol. 2400, (2002). PUBLIC Page 7 of 29
2. MIGRATING DESKTOP PLATFORM 2.1. OVERVIEW The Migrating Desktop Platform is a powerful and flexible user interface, similar to a window-based operating system that hides the complexity of the grid middleware and makes access to the grid resources easy and transparent. Migrating Desktop was created in the CrossGrid project and developed in the int.eu.grid project, where functionality for handling interactive grid applications was implemented. It also proved its usefulness in everyday work of BalticGrid project user s communities chosen as a common point for accessing and managing the project applications, tools, resources and services. Currently it is developed in the BalticGrid-II. It has been also accepted for EGEE RESPECT program. The most important features developed for this prototype were related to support of advanced job types glite parametric jobs and applications based on flow of results (so called workflows). 2.2. PRODUCT DESCRIPTION Name / Acronym: Migrating Desktop Platform (MD) Version: 6.0.11 Main features: Simple integration with grid applications Easy job defining, submission, monitoring and visualization of results Support for batch and interactive jobs Intuitive management of grid data Easy extendable framework Release main features: Handling glite parametric jobs in Migrating Desktop Adaptation of mechanisms used for definition and management of task flow Improvements of advanced data management in Migrating Desktop Integration of Migrating Desktop and VizLitGrid PUBLIC Page 8 of 29
Supported Platforms (OS/Architecture/Grid Middleware): Migrating Desktop Solaris Linux Windows (systems that provide Java >= 1.6 support) Roaming Access Server Scientific Linux CERN SLC release 4.8 glite 3.1 UI License: CrossGrid license 2.2.1. Source code Source code is available through the CVS repository. Release s CVS TAG is: v_6_0_11 Source code repository: 2.2.2. Binary release http://scilla.man.poznan.pl/cgi-bin/viewcvs.cgi/ In addition to source release, MD/RAS is available as a binary release as well. Binary release is compiled with JDK 1.6 and was made available through the RPM packages. RPM packages can be obtained from the RPM repository located at the scilla.man.poznan.pl RPM repository: http://scilla.man.poznan.pl/rpm/6.0.11 RPM packages: http://scilla.man.poznan.pl/rpm/6.0.11/i2glogin-1.1.7-0.i686.rpm http://scilla.man.poznan.pl/rpm/6.0.11/i2g-ras-openldap-1.0-1.i386.rpm http://scilla.man.poznan.pl/rpm/6.0.11/i2g-md-client-6.0.11.noarch.rpm http://scilla.man.poznan.pl/rpm/6.0.11/ i2g-ras-server-6.0.11.noarch.rpm http://scilla.man.poznan.pl/rpm/6.0.11/ i2g-ras-ldap-6.0.11.noarch.rpm http://scilla.man.poznan.pl/rpm/6.0.11/ i2g-ras-server-tomcat6-6.0.11.noarch.rpm http://scilla.man.poznan.pl/rpm/6.0.11/ i2g-ras-mysql-6.0.11.noarch.rpm Dependencies (building and runtime): Migrating Desktop (client) Client side - Java >= 1.6 Server side httpd Roaming Access Server PUBLIC Page 9 of 29
httpd i2glogin i2g-ras-openldap sed /opt/openldap/etc/openldap/slapd.conf mysql >= 4.1.20 mysql-server >= 4.1.20 2.2.3. Documentation and manuals Detailed documentation will be provided with final release of the product and its preparation (or update) was not intended in this phase of development. However the documents concerning MD that covers most important topics: manual, installation guide, tutorial materials can be obtained from the MD s wiki pages (http://scilla.man.poznan.pl:8080/confluence/display/md/migrating+desktop_) or product web pages (http://desktop.man.poznan.pl) Plug-in development: User s guide: http://desktop.psnc.pl/docs/plugins/pluginguide.doc http://support.balticgrid.org/wiki/index.php/migrating_desktop Install guide: http://scilla.man.poznan.pl:8080/confluence/display/euforia/ras+installation+%28scientific+linux+ 4%29 2.3. DESCRIPTION OF WORK Support of advanced job types was a main goal of Migrating Desktop development. Implementation of glite parametric jobs handling required changes in job definition, submission and tracking mechanisms. To make possible also submission of applications based on flow of results (so called workflows), after detailed evaluation of existing tools, the Kepler workflow orchestration system was adopted and tuned for BalticGrid applications. 2.3.1. Handling glite parametric jobs in Migrating Desktop Possibility of submitting of set of jobs differing only by values of input parameters (so called parametric job ) was recognized as one of features required by users. Implementation of such functionality requires changes in job definition, submission and tracking mechanisms. o Changes in job description structures New parameters have been added to internal Migrating Desktop job description structures to support parametric jobs. o Providing support for parametric jobs in submission mechanisms Necessary modification in job submission mechanism has been made to handle parameters of the job. Changes include e.g.: creation of JDL and description of job files. o Job monitoring mechanisms improvements The Migrating Desktop job tracking mechanism was based on command line wrappers, which are not sufficient to support collections of the jobs. To solve this problem new mechanism based on Logging and Bookkeeping java API has been created and integrated with Migrating Desktop server. PUBLIC Page 10 of 29
o Providing support for list of parameters Possibility of defining a list of values as a set of parameters. Necessary modification were made in both: server and client side of Migrating Desktop. o Handling parametric job outcome Output files of sub-jobs of parametric job are now copied to automatically created separated folder on user chosen SE. 2.3.2. Adaptation of mechanisms used for definition and management of task flow Previous version of Migrating Desktop allowed only submitting separate, independent applications. Lack of possibility of defining compound applications based on flow of results between separate jobs was one of gaps identified during BalticGrid-I works. Implementation of workflow mechanism from scratch significantly exceeds JRA activity resources. Keeping in mind requirements of BalticGrid project, after detailed evaluation of existing tools, the Kepler workflow orchestration was chosen. Kepler is a free-software system for designing, executing, evolving, archiving, and sharing scientific workflows (directed graphs where the nodes represent discrete computational components, and the edges represent paths along which data and results can flow between components). Especially important from BalticGrid users point of view are Kepler extensions developed within EU Euforia project, that integrated this platform with glite and UNICORE and provides Kepler as alternative client of Migrating Desktop server. 2.3.3. Improvements of advanced data management in Migrating Desktop Some improvements were made to Grid Commander, a file manager extracted as a separate tool from Migrating Desktop: codes were evaluated and changed to increase stability and efficiency. An access time to LFC files was also optimized 2.3.4. Integration of Migrating Desktop and VizLitGrid VizLitGrid visualization system was integrate within Migrating Desktop platform. Now, a user can use it for visualization of HDF5 directly from Migrating Desktop graphical client 2.3.5. Training materials To let potential users to get familiar with Migrating Desktop a tutorial showing basics of practical Migrating Desktop usage and user guide were prepared and published on Migrating Desktop wiki pages. This work was done to support educational process of SA3 and NA2 activities. 2.4. PLANS FOR THE NEXT RELEASE Migrating Desktop Platform functionality will be continuously extending with a special emphasis put on evaluation, testing and optimization of software, which will lead to final (quality) release. An effort will be focused also on preparation of component documentation (description, user guides, tutorials, etc). PUBLIC Page 11 of 29
3. GRIDCOM 3.1. OVERVIEW Gridcom is a simple web interface for launching applications on grid. The main differences between gridcom and similar systems are: Gridcom allows group access to the same working space (so representing groupware properties). Gridcom is used to launch applications, not single jobs. Each application can use as many jobs as it is needed. Application here means the specific research area, where group of researchers work or group of related jobs for this area. 3.2. PRODUCT DESCRIPTION Name / Acronym: Gridcom Version: 0.99 Main features: Group workspace access Multi-job applications Very simple and native interface Automatic big input file transfer through a SE Automatic aborted job resubmission Highly customizable applications, which can perform any task, like input data scattering, output data collection Release main features: User access control User management interface Compatibility with BalticGrid certificate policy System load optimization Gridcom official web site Documentation Supported Platforms (OS/Architecture/Grid Middleware): Scientific Linux 4, glite License: PUBLIC Page 12 of 29
3.2.1. Source code Gridcom is written in PHP and bash scripts, so its release includes all the source code. 3.2.2. Binary release Latest version: http://sig.balticgrid.org/gridcom/download/gridcom-0.99.tar.gz Dependencies (runtime): 1. Apache HTTP Server Also called HTTPD. Any version. It can be installed anywhere, however, Gridcom will create system HTTPD configuration file /etc/httpd/conf/gridcom.conf, so it s more convenient to have default Red-Hat paths. It will also require some standard modules, like mod_rewrite, mod_proxy, mod_auth atc. 2. PHP Version. System HTTPD has to be configured to handle PHP pages. 3. glite Any version which supports glite-wms-* commands. The system has to be configured to support glite command right after user login. On sig.balticgrid.org this is done, for example, by a simple script /etc/profile.d/glite.sh: if [ "$TERM" ] && [ "`whoami`"!= 'root' ] then echo source /glite3-ui/etc/profile.d/grid_env.sh source /glite3-ui/etc/profile.d/grid_env.sh echo export LFC_HOST=lfc.balticgrid.org export LFC_HOST=lfc.balticgrid.org echo export LCG_GFAL_VO=balticgrid export LCG_GFAL_VO=balticgrid fi 3.2.3. Documentation and manuals Installation instructions: 1. Extract the content of gridcom-0.99.tar.gz to root directory. 2. Ensure that all dependencies are fully satisfied. 3. Add the following two lines into /etc/httpd/conf/httpd.conf file: RewriteEngine On Include /etc/httpd/conf/gridcom.conf If the HTTPD is configured and has HTTPD virtual server, add the lines inside that virtual server configuration. 4. Reload HTTPD configuration by executing `service httpd reload` command. 5. Create new Gridcom for user <username> by executing `gridcom-system-createuser <username>` command. PUBLIC Page 13 of 29
6. Test newly created Gridcom instance by opening in a Web browser URL http://localhost/gridcom/<username>/ or https://localhost/gridcom/<username>/ username: <username> password: thepass 7. User s password can be changed by using command htpasswd -c /gridcom/<username>/httpd/.htpasswd <username> or using user management interface ( Settings menu). Gridcom product documentation is described below. 3.3. DESCRIPTION OF WORK 3.3.1. User access control Now every Gridcom instance can have many read-only users. They can read application and grid log files, study and download application results but can not launch applications on the grid. This is useful for sharing the results and for troubleshooting as administrators can look for any required log file themselves. 3.3.2. User management interface The user management interface allows managing Gridcom users, changing passwords. The interface is accessible through the Settings menu. 3.3.3. Compatibility with BalticGrid certificate policy There was a violation of BalticGrid certificate policies as Gridcom allowed group access to the same Gridcom instance which used one certificate. Now only certificate owner has access to the grid resources, other users have read-only access to the results. 3.3.4. System load optimization The load of Gridcom system server has been drastically minimized by decreasing the rate of checking job statuses and by making this rate adaptive (depends on the count of jobs in an application). 3.3.5. Gridcom official web site The Gridcom homepage (http://sig.balticgrid.org/gridcom) has been created. The site contains documentation (a brochure, an article and a demonstration movie), an informational page for new users with instructions and contacts, and a download page. 3.3.6. Documentation A brochure and an article have been written as part of Gridcom documentation. The article is still under the development, however, it is the main Gridcom documentation source. The latest documentation can be found at Gridcom web site in documentation section (http://sig.balticgrid.org/gridcom/docs). The site also contains a demonstration movie for users. PUBLIC Page 14 of 29
3.4. PLANS FOR THE NEXT RELEASE The next release will have full documentation. PUBLIC Page 15 of 29
4. PARAVIEW 4.1. OVERVIEW The most powerful open-source visualization software ParaView [1] has been deployed in BalticGrid- II in order to provide for grid users enhanced application services aimed for distributed visualization of large datasets. ParaView is very useful for BalticGrid users performing large distributed computations of actual industrial problems like oil filters, sediment transport, compacting and hopper discharge. Visualized applications belong to the BalticGrid-II application group named Engineering Modelling Tasks. ParaView is an open-source application designed to visualize large datasets. ParaView supports hardware-accelerated parallel rendering and achieves interactive rendering performance via level-ofdetail (LOD) techniques. ParaView runs on distributed and shared memory parallel machines as well as single processor PC and has been successfully tested on Windows, Mac OS X, Linux and various Unix workstations, clusters and supercomputers. Under the hood, ParaView uses the VTK [2] as the data processing and rendering engine. ParaView provides highly configurable GUI based on Qt [3] for the interactive exploration of large datasets. One of the most convenient ways to automate ParaView is to use the Python scripting [4]. ParaView comes with a program called pvbatch. It can run in parallel without having to establish a slow client-server connection. Grid user can start interactive visualization session employing full power of GUI or use a batch mode producing final images and animations. Two modes of user interaction have been implemented and deployed. Interactive mode allows for users to employ full power of GUI and highly interactive widgets. Interactive GUI can run on grid by using client-server communication mode named reverse connection. ParaView program pvserver, containing data server and render server, runs on WNs and establishes a reverse socket connection between the first node of the server (0-th MPI node) and the client running on a local PC. Alternative batch mode is very attractive for experienced grid users that generate long animations. GUI is disabled and the slow socket communication is replaced by usual grid protocols and utilities for file transfer. 4.2. PRODUCT DESCRIPTION Name / Acronym: ParaView Version: 3.6.1 Main features: Extensive visualization capabilities handling structured, unstructured, polygonal, image, multi-block and AMR data types. Intuitive and flexible interface based on the Qt application framework. Supports a variety of file formats including: VTK, EnSight, STL and BYU. Runs parallel on distributed and shared memory systems using MPI. ParaView is fully scriptable using the simple but powerful Python language. Additional modules can be added by either writing an XML description of the interface or by writing C++ classes. Interactive work on grid employing full power of GUI. PUBLIC Page 16 of 29
Release main features: Implemented parallel HDF5 reader. GPU rendering employed on grid. Supported Platforms (OS/Architecture/Grid Middleware): Operating Systems: Linux x86, 32 bit and 64 bit Windows, 32 bit and 64 bit Tested interoperability with grid middleware: glite 3.1.0 License: Kitware open-source license 4.2.1. Source code Source code of ParaView 3.6.1 can be downloaded from official web page: http://www.paraview.org/files/v3.6/paraview-3.6.1.tar.gz Alternative location for software download is provided on the SVN repository: http://grid.vgtu.lt/svn/downloads/paraview-3.6.1.tar.gz Software necessary for compilation from the source code can be also downloaded from: Cmake 2.6.4: cmake-2.6.4-linux-i386.tar.gz Qt 4.3.0: http://grid.vgtu.lt/svn/downloads/qt-all-opensource-src-4.3.0.tar.gz OpenMPI 1.3.3: openmpi-1.3.3.tar.gz Mesa 7.5.1: MesaLib-7.5.1.tar.gz Python 2.3.4: http://www.python.org/download/ Deployment of the ParaView server on grid is based on compilation of the source code. It supports MPI, but does not support GUI. The SGM scripts for automatic ParaView (employing CPU rendering) installation can be obtained from: http://grid.vgtu.lt/svn/griddeployment/a-graphical-paraview- 3.6.1/ParaviewOffscreenInstall http://grid.vgtu.lt/svn/griddeployment/a-graphical-paraview- 3.6.1/ParaviewOffscreenSetup.jdl The SGM scripts for automatic ParaView (employing GPU rendering) installation can be obtained from: http://grid.vgtu.lt/svn/griddeployment/a-graphical-paraview-3.6.1/paraviewdriinstall http://grid.vgtu.lt/svn/griddeployment/a-graphical-paraview-3.6.1/paraviewdrisetup.jdl Developed parallel reader for datasets stored in predefined HDF5 format [5] is provided as plug-in. In order to employ this special purpose reader, ParaView client and server should be compiled and build from the source. The source code of the reader can be downloaded from SVN repository: PUBLIC Page 17 of 29
http://grid.vgtu.lt/svn/downloads/hdfreaders.tar.gz 4.2.2. Binary release ParaView 3.6.1 software is available as a binary release as well. It can be useful for installation of the ParaView client on user PC. Pre-compiled binaries can be downloaded from official web page: Windows, 32 bit: http://www.paraview.org/files/v3.6/paraview-3.6.1-win32-x86.exe Windows, 64 bit: http://www.paraview.org/files/v3.6/paraview-3.6.1-win64-x86.exe Linux x86, 32 bit: http://www.paraview.org/files/v3.6/paraview-3.6.1-linux32-x86.tar.gz Linux x86, 64 bit http://www.paraview.org/files/v3.6/paraview-3.6.1-linux64-x86.tar.gz Dependencies (Compilation of the source code): Cmake >= 2.6.4 gcc >= 3.4.6 OpenMPI 1.3.3 Mesa or OpenGL Xorg-X11-devel Qt >= 4.3 Python 4.2.3. Documentation and manuals Documentation on ParaView 3.6.1 deployment in grid covers several important topics: SGM installation, Grid User s Manual including tutorial and testing example. All documents can be obtained from the web page of ParaView in BalticGrid-II: Web page: http://lsl.vgtu.lt/paraview/ SGM Installation Guide: http://lsl.vgtu.lt/userfiles/paraview_sgm_installationguide_2009.pdf Grid User s Manual: http://lsl.vgtu.lt/userfiles/paraview_gridusersmanual_2009.pdf Testing example of developed HDF5 reader: http://grid.vgtu.lt/svn/downloads/paraview_hdf5reader_example.tar.gz Full ParaView documentation can be found on the official web site and wiki pages: http://www.paraview.org/paraview/help/documentation.html http://paraview.org/wiki/paraview PUBLIC Page 18 of 29
4.3. DESCRIPTION OF WORK In [6] presented performance analysis reveals that parallel speed-up is not increased reading data by automatic readers. Moreover, the data decomposition performed by vtkdistributeddatafilter can be longer than the attained improvement in work of the visualization pipeline. Special purpose parallel reader is developed for unstructured datasets stored in predefined HDF5 format. It is adapted for the nature of pilot applications that decompose the solution domain into subdomains, each being assigned to a processor, and ensure load balancing. Each process writes its own result file, which can be read and visualised independently. Developed reader is based on the idea that each process independently reads its file, performs visualization and sends resulting image to the 0-th MPI node. Parallel rendering library IceT employs a sort-last algorithm, composites the partial images together and forms the final image. Information on decomposed data files is stored in XML file, which is read by main process running on the 0-th MPI node. Each process receives the name of data file and other parameters from the main process. Then it works independently until partial image is send to the main process. Efficiency of performed visualization is completely insensitive to the amount of data being rendered. This makes it a very scalable algorithm well suited to large data. ParaView version 3.6.1 has been deployed in BalticGrid-II testbed for visualization. In this deployment, GPU or direct rendering can be employed for visualization as well as CPU rendering. It is worth to note, that glite JDL abilities are not enough flexible for running parallel MPI jobs on multicore architectures. Thus, GPU rendering can not be directly applied on multi-core architectures. Two processes running on two cores of the same node can employ the same video RAM for GPU rendering and produce artefacts in the final image. ParaView server should be carefully configured in order to obtain right image and avoid artefacts. Deployment has been performed by using BalticGrid-II SGM (Software Grid Manager) system. The ParaView tags are VO-balticgrid-A-GRAPHICAL-PARAVIEW-3.6.1-OSMESA (CPU rendering) and VO-balticgrid-A-GRAPHICAL-PARAVIEW-3.6.1-DRI for direct rendering (GPU rendering). Required environment can be read from the source file env.sh. Installed software resides in the predefined directories $VO_BALTICGRID_SW_DIR/A-GRAPHICAL-PARAVIEW-3.6.1-OSMESA and $VO_BALTICGRID_SW_DIR/A-GRAPHICAL-PARAVIEW-3.6.1-DRI. The SGM scripts for automatic software installation has been prepared and located in the SVN repository. Some ParaView dependencies like Mesa and Open MPI have been installed by using separate SGM scripts. ParaView has been tested on several sites that are marked by ParaView tag: ce2.grid.vgtu.lt ce.grid.vgtu.lt ce.cyf-kr.edu.pl Direct rendering (GPU rendering) is only supported in ce2.grid.vgtu.lt, because high performance GPU (Nvidia GeForce 9600GT 512MB 256bit) are available on this grid cluster. PUBLIC Page 19 of 29
5. GRID VISUALIZATION E-SERVICE VIZLITG 5.1. OVERVIEW Grid visualization e-service VizLitG is designed for convenient access and interactive visualization of remote data files located in grid. The software is developed in GirdTechno project supported by the Lithuanian State Science and Studies Foundation. VizLitG is adapted to needs of BalticGrid users and employed forr visualization of scientific results computed in BalticGrid [7]. The client-server architecture of VizLitG is based on widely recognized web standards. The e- service is implemented in GlassFish application server [8]. The visualization server runs on special User Interface for Graphics (UIG). Thus, natural access to grid resources is available. The lightweighted client implemented as Java application consists of GUI and Viewer. Metadata, visualization network and its parameters are transferred by JAX-WS runtime while a final image and interactivity is provided by GVid [9]. Data files located in SE can be accessed by using traditional means like LFC (LCG) available in glite distribution [10]. The additional application server runs on the experimental SE in order to provide special data services able processing and transferring the selected parts of datasets. In order to provide for remote users high interactivity level, GVid software was implemented in VizLitG as video-streaming module. The most important GVid classes were renewed to support VTK 5.4. and enwrapped by Java. Thus, remote user has full interactivity provided by GVid and VTK widgets. GUI allows automatic data management and interactive dataset selection. In order to process HDF5 files automatically, datasets are stored in predefined structure allowing the software to interpret the structure and contents of a file without any outside information. HDF5 groups and datasets are automatically processed considering values of HDF5 attributes. Time dependent and time independent data are processed differently. Datasets that do not change in time are located in the group Common. Datasets varying in time are grouped and stored according to time step number. GUI separates geometry and topology from attributes like scalars or vectors in order emphasize their different nature. The visualization engine of the VizLitG is based on VTK toolkit. VTK objects are enwrapped by Java programming language in order to build the service. A visualization network is assembled from VTK objects by using GUI. The resulting pipelines are described by XML language. Valid XML documents are automatically generated on a client and transferred to the server by JAX-WS. Tabular design of GUI fields simply illustrates the dataflow. 5.2. PRODUCT DESCRIPTION Name / Acronym: VizLitG Version: 0.3 Main features: Automatic HDF5 data management and interactive data selection based on GUI, VTK visualization network automatically described by XML, Visualization of data files residing locally on UIG, Visualization of full data files transferred from SE, Visualization of interactively selected parts of datasets transferred from SE. PUBLIC Page 20 of 29
Release main features: Revisited security, Enhanced interactivity including VTK widgets, Supported Platforms (OS/Architecture/Grid Middleware): Client: Linux x86 Windows, 32 bit Server: Scientific Linux SL release 5.2 (Boron) Tested interoperability with grid middleware: glite 3.1.0 License: GridTechno open-source license 5.2.1. Source code Source code of VizLitG and related software packages can be downloaded from the SVN repository. The source code of visualization server: Main service on UIG: http://grid.vgtu.lt/svn/downloads/vizlitgserv.tar.gz Visualization engine: http://grid.vgtu.lt/svn/downloads/vizsystem.tar.gz Data service on SE: http://grid.vgtu.lt/svn/downloads/serverse.tar.gz Maintaining software for compilation of visualization server from the source code can be also downloaded: GlassFish v2: http://java.sun.com/javaee/downloads/index.jsp VTK 5.4.2: http://grid.vgtu.lt/svn/downloads/vtk-5.4.2.tar.gz Renewed VTK-GLUT: http://grid.vgtu.lt/svn/downloads/vtk_mod.tar.gz GVid: http://grid.vgtu.lt/svn/downloads/gvid.tar.gz HDF5 1.8.1: http://grid.vgtu.lt/svn/downloads/hdf-java.tar.gz Grid security libraries: http://grid.vgtu.lt/svn/downloads/security-libs.tar.gz Glassfish security library: http://grid.vgtu.lt/svn/downloads/webservices-rt.jar XML parser: http://grid.vgtu.lt/svn/downloads/jdom.jar The source code of the client: Visualization client: http://grid.vgtu.lt/svn/downloads/vizlitg.tar.gz The client needs maintaining software: Glassfish security library: http://grid.vgtu.lt/svn/downloads/webservices-rt.jar XML parser: http://grid.vgtu.lt/svn/downloads/jdom.jar PUBLIC Page 21 of 29
5.2.2. Binary release The visualization client of the e-service VizLitG can run on any PC with any Linux OS or Windows OS. The client software can be automatically downloaded and installed by Java Web Start [11] application from URL address: http://lsl.vgtu.lt/soft/vizlitg/launch.jnlp Binary release of visualization server is not available, because the compiled software is highly dependent on graphics hardware. Dependencies (building and runtime): Visualization server: Java >= 1.6 NetBeans IDE >= 6.1 GlassFish v2 VTK 5.4.2 Mesa or OpenGL GVid HDF5 1.8.3 Visualization client: Java >= 1.6 5.2.3. Documentation and manuals Documentation of VizLitG covers several important topics: server installation, user manual, short tutorial and testing example. All documents can be found in the web page of VizLitG: Web page: http://lsl.vgtu.lt/vizlitg Server Installation Guide: http://lsl.vgtu.lt/userfiles/vizlitg_installationguide_2009.pdf User s Manual: http://lsl.vgtu.lt/userfiles/vizlitg_gridusersmanual_2009.pdf Grid Tutorial: http://lsl.vgtu.lt/userfiles/vizlitg_griduserstutorial_2009.pdf Dataset for testing: http://grid.vgtu.lt/svn/downloads/ubgrid_1_1.h5 5.3. DESCRIPTION OF WORK New version 0.3 of the visualization e-service VizLitG has been deployed in BalticGrid-II. In order to increase availability of the e-service, Java based GVid SDL output client developed at g-eclipse project was implemented into VizLitG. Thus, the complete set of the client software can be automatically downloaded and installed by Java Web Start technology on Linux OS as well as on Windows OS. The Message Authentication over SSL mechanism of GlassFish is employed for security reasons. The server authenticates a client of the visualization service by using basic authentication scheme including name/password pair of credentials. Moreover, HTTPS protocol using Secure Sockets Layer PUBLIC Page 22 of 29
(SSL) is used for message integrity and confidentiality. Grid user authentication and authorization is performed by using traditional grid proxy certificate and glite means in order to access grid resources like data located in SE. The complete set of UI software is installed on UIG, therefore, visualization service is able to perform user authorization and to access grid resources. The interactivity of e-service was enchanted by implementing VTK widgets. Now users can interactively select parts of the resulting visualization or interactively define geometric objects that will be processed by following visualization filters. Finally, the VizLitG was adapted to the new VTK version 5.4.2. Deployment of e-service is quite different from that of usual visualization software or any other grid application. VizLitG is closely related with glite middleware and grid infrastructure like storage elements, but current software implementation does not require any computing resources. The visualization server or UIG is deployed on HP xw4600 workstation empowered by Nvidia Quadro FX4600 (768MB) for GPU rendering: uig.grid.vgtu.lt Currently, the e-service is not supposed to be deployed on other servers. The source codes of visualization server are provided only for consistency reasons. The service providing the access of interactively selected parts of datasets is implemented in the experimental SE: se.grid.vgtu.lt Maintenance of e-services and hosting hardware is provided by VGTU SA1 team. User support is provided by VGTU SA3 team. User s Manual and Grid Tutorial have been prepared and have been placed on the web page of the e-service VizLitG. 5.4. PLANS FOR THE NEXT RELEASE VizLitG is being developed in the NGI project, therefore, it is difficult to plan the next software release. However, deployment of VizLitG in BalticGrid-II encourages constructive cooperation with Lithuanian NGI. During the end of the year main development efforts are targeted to computational steering or real time visualization of running computations. This feature should be available in January of 2010. PUBLIC Page 23 of 29
6. SPECIALIZED VISUALIZATION TOOL FOR PARTICLE SYSTEMS VISPARTDEM 6.1. OVERVIEW VisPartDEM is distributed visualization tool for particle systems simulated by the Discrete Element Method. Discrete element computations are based on particle positions, forces acting between particles and Newton s laws. Particle systems have no permanent connections or usual grid that can be applied for spatial discretization or visualization purposes. Initial defects are identified between pairs of neighbouring particles that can only form 1D connections in 2D or 3D space. The resulting physical phenomenon has no continuous structure. 1D connections are not suitable for reliable interpolation and standard visualization. Intensive post-processing stage is necessary for visualization of derived variables of interest. Advanced algorithms based on surface (curve in 2D) extraction are implemented in order to obtain geometric representation of propagating cracks. Distributed visualization tool is based on data parallel model of VTK [2] and distributed parallel rendering of IceT. Visualization engine of VisPartDEM runs on working nodes of any computing clement. VisPartDEM employs GVid software [9] as video-streaming module. As a result VisPartDEM runs on WNs while the compressed video stream is efficiently transferred through the network from 0-th MPI node and displayed on client. Remote grid user has full interactivity provided by VTK widgets and interactor styles. 6.2. PRODUCT DESCRIPTION Name / Acronym: VisPartDEM Version: 0.1 Main features: Visualization pipeline described by XML. Parallel data model of VTK, Distributed parallel rendering, Full interactivity level on grid provided by GVid. Release main features: Distributed parallel visualization of large particle systems, Visualization of forces acting between particles, Visualization of defects between pairs of neighbouring particles, Special functionality for crack detection and visualization. Supported Platforms (OS/Architecture/Grid Middleware): Client: Linux x86, 32 bit Windows, 32 bit Server: Linux x86, Scientific Linux SL release 5.2 (Boron) PUBLIC Page 24 of 29
CE and WNs Linux x86, 32 bit Tested interoperability with grid middleware: glite 3.1.0 License: Open-source license 6.2.1. Source code VisPartDEM consists of several parts: client running on user PC, visualization engine running on working nodes of CE, main service running on UIG server and data management service running on experimental SE. Source code of VisPartDEM and related software packages can be downloaded from the SVN repository. The source code of visualization server: Main service on UIG: http://grid.vgtu.lt/svn/downloads/vizpartdemserv.tar.gz Visualization engine: http://grid.vgtu.lt/svn/downloads/vizpartdem_engine_src.tar.gz Data service on SE: http://grid.vgtu.lt/svn/downloads/serverse.tar.gz Maintaining software for compilation of visualization server from the source code can be also downloaded: GlassFish v2: http://java.sun.com/javaee/downloads/index.jsp HDF5 1.8.3: http://grid.vgtu.lt/svn/downloads/hdf-java.tar.gz Grid security libraries: http://grid.vgtu.lt/svn/downloads/security-libs.tar.gz Glassfish security library: http://grid.vgtu.lt/svn/downloads/webservices-rt.jar XML parser: http://grid.vgtu.lt/svn/downloads/jdom.jar The source code of the client: Visualization client: http://grid.vgtu.lt/svn/downloads/viz_part_dem.tar.gz The client needs maintaining software: Glassfish security library: http://grid.vgtu.lt/svn/downloads/webservices-rt.jar XML parser: http://grid.vgtu.lt/svn/downloads/jdom.jar Visualization engine is installed in grid by using Software Grid Manager system. VTK forms the basis of visualization engine of VisPartDEM. Source code of VTK 5.4.2 can be downloaded from official web page: VTK 5.4.2: http://www.vtk.org/files/release/5.4/vtk-5.4.2.tar.gz Alternative location is provided on SVN repository: PUBLIC Page 25 of 29
VTK 5.4.2: http://grid.vgtu.lt/svn/downloads/vtk-5.4.2.tar.gz Related software for compilation of visualization engine can be also downloaded: Cmake 2.6.4: http://grid.vgtu.lt/svn/downloads/cmake-2.6.4-linux-i386.tar.gz GVid: http://grid.vgtu.lt/svn/downloads/gvid.tar.gz OpenMPI 1.3.3: http://grid.vgtu.lt/svn/downloads/openmpi-1.3.3.tar.gz Freeglut 2.6.0: http://grid.vgtu.lt/svn/downloads/freeglut-2.6.0-rc1.tar.gz SDL 1.2.13: http://grid.vgtu.lt/svn/downloads/sdl-1.2.13.tar.gz Xvidcore 1.2.2: http://grid.vgtu.lt/svn/downloads/xvidcore-1.2.2.tar.gz Renewed VTK-GLUT: http://grid.vgtu.lt/svn/downloads/vtk_mod.tar.gz HDF5 1.8.3: http://grid.vgtu.lt/svn/downloads/hdf5-1.8.3.tar.gz Deployment of the VisPartDEM on grid is based on compilation of the source code and VTK including several GVid classes. The SGM script for automatic software installation can be obtained from: Some VisPartDEM dependencies like Mesa v7.5.1, OpenMPI v1.3.3 and HDF5 v1.8.3 can be installed by using separate SGM scripts: MesaInstall: http://grid.vgtu.lt/svn/griddeployment/a-graphical-mesa-7.5.1/mesainstall MesaSetup.jdl: http://grid.vgtu.lt/svn/griddeployment/a-graphical-mesa- 7.5.1/MesaSetup.jdl OpenMpiInstall:http://grid.vgtu.lt/svn/gridDeployment/E-PARALLEL-OPENMPI- 1.3.3/OpenMPIInstall OpenMpiSetup.jdl:http://grid.vgtu.lt/svn/gridDeployment/E-PARALLEL-OPENMPI- 1.3.3/OpenMPISetup.jdl Hdf5Install: http://grid.vgtu.lt/svn/griddeployment/e-devel-hdf5-1.8.3/hdf5install Hdf5Setup.jdl: http://grid.vgtu.lt/svn/griddeployment/e-devel-hdf5-1.8.3/hdf5setup.jdl Two VTK 5.4.2 implementations are considered. The first one is based on CPU rendering, while the second one employs GPU for direct rendering. VTK SGM scripts can be downloaded from SVN repository. CPU rendering is employed: VtkGvidOSMesaInstall: http://grid.vgtu.lt/svn/griddeployment/a-graphical-gvid-vtk- 5.4.2/VtkGvidOSMesaInstall VtkGvidOSMesaSetup.jdl: http://grid.vgtu.lt/svn/griddeployment/a-graphical-gvid-vtk- 5.4.2/VtkGvidOSMesaSetup.jdl GPU rendering is employed: VtkGvidDRIInstall: http://grid.vgtu.lt/svn/griddeployment/a-graphical-gvid-vtk- 5.4.2/VtkGvidDRIInstall VtkGvidDRISetup.jdl: http://grid.vgtu.lt/svn/griddeployment/a-graphical-gvid-vtk- 5.4.2/VtkGvidDRISetup.jdl PUBLIC Page 26 of 29
6.2.2. Binary release The visualization client of VisPartDEM can run on any PC with any Linux OS or Windows OS. The client software can be automatically downloaded and installed by Java Web Start application from URL address: http://lsl.vgtu.lt/soft/vispartdem/launch.jnlp Binary release of the visualization engine is not available, because the compiled software is highly dependent on graphics hardware. Dependencies (building and runtime): Cmake 2.6.4 gcc 3.4.6 OpenMPI 1.3.3 HDF5 1.8.3 Mesa or OpenGL Freeglut 2.6.0 SDL 1.2.13 Xvidcore 1.2.2 6.2.3. Documentation and manuals Documentation on VisPartDEM deployment in grid covers several important topics: Installation, Grid User s Manual including tutorial and testing example. All documents can be obtained from the web page of VisPartDEM: Web page: http://lsl.vgtu.lt/vizpartdem/ Installation Guide: http://lsl.vgtu.lt/userfiles/vispartdem_installationguide.pdf Grid User s Manual: http://lsl.vgtu.lt/userfiles/vispartdem_gridusersmanual.pdf Testing example: http://grid.vgtu.lt/svn/downloads/tension_plokst_1.h5 6.3. DESCRIPTION OF WORK Initial version of visualization tool for particle systems described in DJRA1.3 has been reconstructed and run in parallel. Data parallel model of VTK has been employed for visualization of large particle systems. Distributed parallel rendering based on IceT has been integrated with GVid software in order to transfer images and interactivity from 0-th MPI node to the client. Parallel HDF5 reader has been incorporated into VisPartDEM and has been applied for independent reading of decomposed poly data files. Advanced filters for visualization of propagating cracks and safety margin forces have been implemented in the VisPartDEM software. Visualization of considered particle-systems needs time consuming post-processing and extended functionality based on feature extraction. Several algorithms for crack visualization have been implemented. The simplest algorithm maps initial defects to generated triangles (tetrahedras). Spatial representation of cracks can be visualized as cell attributes. Advanced algorithms employ global or local Voronoi diagrams in order to obtain contact surfaces of PUBLIC Page 27 of 29
neighbouring particles. Finally, cracks are represented by colored tubes plotted on contact surfaces. Attention is also focused on visualization of forces acting between particles like safety margin force. Such forces can be thresholded and represented as colored points, colored glyphs or colored tubes. Other visualization filters and mappers have been implemented in the software: Clipping, Extract geometry, Threshold, Particle coloring by scalar (radius, processor ID), Vector visualization by glyphs (velocities, accelerations, forces), Advanced functionality and standard filters for VisPartDEM is continuously enriched and tested. GUI of VisPartDEM is implemented as e-service in GlassFish application server. It allows interactive browsing of SE content ant automatic data management. The GUI is light-weighted client implemented as Java application. Metadata, visualization network and its parameters are transferred by JAX-WS runtime while a final image and interactivity is provided by GVid. VisPartDEM 0.1 based on VTK 5.4.2 has been deployed in BalticGrid-II testbed for visualization. Deployment has been performed by using BalticGrid-II SGM (Software Grid Manager) system. Software has been installed in the predefined location, which is specified by contents of $VO_BALTICGRID_SW_DIR variable and VisPartDEM tag VO-balticgrid-A-GRAPHICAL-GVID- VTK-5.4.2-OSMESA (CPU rendering) and VO-balticgrid-A-GRAPHICAL-GVID-VTK-5.4.2-DRI for direct rendering (GPU rendering). Required environment can be read from the source file env.sh. Installed software resides in the predefined directories $VO_BALTICGRID_SW_DIR/A-GRAPHICAL- GVID-VTK-5.4.2-OSMESA and $VO_BALTICGRID_SW_DIR/A-GRAPHICAL-GVID-VTK-5.4.2-DRI. VisPartDEM has been tested on several sites that are marked by VisPartDEM tag: ce2.grid.vgtu.lt ce.grid.vgtu.lt The dataset for testing can be found on the web page of VisPartDEM. Installation Guide and Grid User s Manual of the software including short tutorial is also available in the same location. PUBLIC Page 28 of 29
7. SUMMARY AND CONCLUSIONS Implementation works of the second stage of the BalticGrid project and development of three main components of the activity: Migrating Desktop, Gridcom and visualization tools result in preparation of the second prototype described in this document. In the next period the main effort will be focused on evaluation, testing and optimization of provided software which will result in preparation of final (quality) release of JRA components (PM24). Very important task will be also preparation of JRA components documentation and support provided to NA2 and SA3 activities (description, user guides, tutorials, etc). PUBLIC Page 29 of 29