ULabGrid, an infrastructure to develop distant laboratories for undergraduate students over a Grid O.Ardiaz, D.Royo et al. Dept. of Computer Architecture Polytechnic University of Catalonia, Spain
Outline Introduction Advantages of using ULabGrid Implementation Architecture Functionality Conclusions and future work plans
Introduction Virtual laboratories –Ease overcrowding –Access from anywhere –Provide operating system services for networked resources –Facilitate necesarry tools –On-line manuals, FAQs, examples etc.
Advantages of using ULabGrid Dedicated software installation and maintenance Reduces licensing problems Extended hardware resource availability Ease of changing lab content Simplified user interfaces
Implementation Uses: –Globus (2.2) –GPDK (Globus Portal Development Kit) –VNC (Virtual Network Computing) Main components –Session scheduler –Resource manager
Architecture Pcribes.ac..upc.es ReplicaMgr MetadataDirect oryServer Grid-Proxy GassServer GlobusPortal User Directory Session Scheduler Resource Mgr GassServer Globus-Gatekeeper File Server GSSHFTP Globus- Gatekeeper avant.upc.es Pentium IV pcmartino.ac.upc.es 1->Deploy Application Student Labs Instructor: Manages Ulab Apps %Vnc-server:1 %nam & GassServer Globus-Gatekeeper Pentium IV pc3.ac.upc.es GassServer Globus-Gatekeeper Pentium IV pc5.ac.upc.es GassServer Globus-Gatekeeper Pentium IV pc4.ac.upc.es GassServer Globus-Gatekeeper Pentium IV pc5.ac.upc.es ULab Grid Web Browser VNC-client 2->Launch NAM 3->Put User Files 4->Exec VNC, Exec NAM 5-Return VNC port 6->Interact with NAM
Functionality 1.Users connect via a browser to portal and choose tool to run, files needed etc. 2.Session Scheduler queries the Resource Manager for optimal machine and transfers needed files (GSIFTP) to it 3.VNCserver process at remote machine started and tool executes, VNC parameters sent to user 4.VNCviewer started by user 5.When done, Session Scheduler does cleanup
GPDK Architecture
Administrator interface
Conclusions and future work plans Propose looking into: –Process migration due to node failure –Resource mapping algorithms –System dynamics – notification services –Collaboration, working in groups