• Goal : Run application-oriented benchmarks to measure the performance of StratusLab cloud implementation. Benchmarks tests will be deployed on StratusLab cloud, results will be stored in xml format.
  • Run cpu_benchmarks.sh for High-CPU requirements but little or no input and output data.
  • Run mpi-io_benchmarks.sh for Simulation, Analysis and Filtering tests
  • Run openmpi_benchmarks.sh for shared Memory Programs
  • Run mpi_benchmarks.sh for MPI-like programs.
  • Run kepler_benchmarks.sh for workflow multiple interdependents tasks. We are using kepler as test platform for workflow orchestration.

• Install a cloud from scratch, using manual tools (/usr/sbin/stratus-*) or automated tools (Quattor)
• Remotely monitoring the status of the cloud (web and /usr/bin/stratus-*)
  • stratus-describe-node to list nodes and show their status
  • stratus-describe-instance to list running vms and show their status
  • Same thing with the web
• Run instances with stratus-run-instance
  • demo contextualization
  • ssh into instance with user private key, over public ip
• Create image with stratus-create-image
  • demo contextualization (works with stratus-run-instance)
  • show result in appliance repository
• Appliance Repository
  • show manifest file and image
  • show squid feature (if available)

• Capability to deploy Grid sites on top of Cloud infrastructures running OpenNebula, using pre-configured virtual appliances with grid middleware (gLite) already installed.
• The virtual appliances repository that provide the pre-configured images.
• Capability to manage virtualized grid sites by adding and removing Worker Nodes on the fly.
• Capability to run various applications and benchmarks
• Capability to monitor the cloud infrastructure using the OpenNebula XML-RPC API.

• A Cloud site with OpenNebula having enough physical hosts in order to be able to support multiple grid sites with enough WNs.
• A gLite UI that support the StratusLab VO
• Graphical monitoring of ONe sites
• The appliances repository with Web front-end that enables browsing and search of appliances and metadata.
• A set if Appliances with pre-configured gLite middleware for CE, SE and WNs. Choice of OpenMPI support should be available.
• A set of command line tools that enable the parametric setup and configuration of grid sites based on templates.
• A set of 2-3 applications/benchmarks that can be used to showcase the efficiency of the virtualized site. One of these applications should be parallel/MPI. The applications should come pre-installed with the WN images.

1. STRATUSLAB-65
2. STRATUSLAB-67
3. STRATUSLAB-68
4. STRATUSLAB-79
5. STRATUSLAB-144
6. STRATUSLAB-145
7. STRATUSLAB-151 (if we want to use the bioinformatics application)
8. STRATUSLAB-154
9. STRATUSLAB-155
10. STRATUSLAB-156

The presenter logs into the OpenNebula frontend and runs a few commands to show that status of the site. He also has a browser which shows the current state of the cloud infrastructure next to the shell window. The presenter runs a command that prepares a complete grid site on-the-fly using a set of site configuration templates. The site is prepared based on the appliances available from the repository. The presenter shows the web site of the repository and explains how it works, what type of appliances are available at the moment and how someone can search for specific appliances based on a set of metadata that are currently being kept. The site offers plenty of storage space in the SE and shared space in WNs for VO applications and as a scratch space for MPI applications.

The presenter runs a sample job to demonstrate the validity of the site. Then he/she registers and uploads sample data file on the SE using the LFC. Next the presenter runs an MPI application and demonstrates how the current solution can be used for scientific applications from various domains.

The presenter alters the configuration of the site by removing and adding WNs by hand. The action has as a result the update of site configuration in the CE. The presenter shows on the graphical monitoring tool the new configuration of the site. He/she runs again the parallel job with the new set of WNs to show that the application can now take advantage of the additional WNs (and potentially finish faster).