Simplifying Grid Complexity with SSH Session Server Framework
Explore SSH Session Server framework for grid computing, CLI models, application managers, interactions, and practical demos. Simplify user experience and enhance productivity without modifying existing applications.
Simplifying Grid Complexity with SSH Session Server Framework
E N D
Presentation Transcript
Hiding Grid Complexity Behind SSH Session Server framework Tomasz Kuczyński (1,2) docentt@man.poznan.pl Piotr Kopta (2) kopta@icis.pcz.pl 1) Poznan Supercomputing and Networking Center 2) Czestochowa University of Technology
Outline • SSH Session Server framework • CLI Model Example • Framework Architecture • Application Managers • Interactions between the components • Demo
SSH Session Server framework • What exactly is SSH Session Server framework ? • Why SSH ? • Installed on each resource • No modifications of existing applications needed • Why PTYs ? • utilization of standard SSH client • How does framework work with SSH session ? • CLI interaction model • Application descriptors
SSH Session Server framework (cont.) • Main goal of the framework is to gain high level of business logic – presentation layer separation • SSH Session Server - continuation of WebCI portal solutions • Features: • Adding user-defined interfaces at portal run-time • Easy adaptation of existing applications • Seamless installation • XML-based application interface description • VRML, X3D, SVG, Charts (jpeg, png) support
Application Managers • Simple applications that have CLI • Application manager using the grid resource management system such as GRMS may submit many parallel jobs (reduce apps run-time) and control all the computing on user’s behalf • Overall system performance is increased • User interaction is limited to required minimum • run a job • view results • do not worry about resources • Apps manager will adjust number of jobs running in parallel to the number of resources available
Interaction between the components • Portal <-> Application Manager <-> Grid Infrastructure • User sets up the input parameters and orders to start an application • Portal passes input data to Application Manager (APPMGR) • Application manager uses the GRMS GAT adaptor to run an application • Resource and job description generated by APPMGR is sent to GRMS • GRMS finds appropriate resources and runs a job • GRMS, using replica management system or GRMS’s own file movement mechanisms, stages the files • GRMS starts a job and returns job handler to the APPMGR
Interactions (cont.) • APPMGR collects all the results and presents them to the portal • User decides about the next steps • Chooses the visualization style • HTML, XML, VRML, X3D, SVG, JFreeChart • Plays with the results and decides about further application runs • Views the history of the application runs • Sets up the notification mechanisms (email, sms, other...) • All user activities are asynchronous and non-blocking • A user can for example start a job and then, while waiting for the results, logout, go for a coffee and then return to the application
Today’s demo • We will show an interactive Heat Transfer application run using GridLab and ClusterIX tools (Portal, GAT, GRMS, GAS, Mobiles Service, other) and SSH Session Server framework • Application user perspective • Grid hidden behind the portal • No job description interface • Why should user struggle with it? This is too complex... • No data transfer interface • No less complex than job description interface, so why bother user about it? • No resource discovery interface • Why should user know anything about the resources he might run the job at?
Questions For more information visit: • GridSphere http://www.gridsphere.org • GridLab http://www.gridlab.org • Open Grid Portals http://www.opengridportals.com • ClusterIX http://www.clusterix.pl Thank you