; downloading
Documents
Resources
Learning Center
Upload
Plans & pricing Sign in
Sign Out
Your Federal Quarterly Tax Payments are due April 15th Get Help Now >>

downloading

VIEWS: 0 PAGES: 16

  • pg 1
									          E-x-t-e-n-d-i-n-g Rocks:
       The Creation and Management of Grid
           Systems for Biomedical Research

            OSGC Conference - May 14, 2008



                                         Vicky Rowley
                                     Solution Architect
BIRN Coordinating Center - University of California San
                                                  Diego
BIRN is Data Storage (SRB) and Processing,
but…


                                    UMN
                                                     MGH
                                           BWH
Stanford                                             Yale
                                      UI
                                           Cluster
UCLA

  UCI                                      Duke
                              UNM
     UCSD
              Cluster



   = Support existing sites
   = Establish new sites
   = Replicate for new community
Rocks Standard vs. Rocks for BIRN
 Cluster building focus    Collaboration focus
 Data processing focus     Data storage/sharing focus
 Lots of big clusters      A few relatively small
                             clusters
                            Data Grid was needed
                             before clusters processing
                             was needed
So what does BIRN _do_?
                                         Large Deformation Diffeomorphic
                                        Metric Mapping using the TeraGrid



                               4        JHU
                                    Shape Analysis
       3
              MGH               of Segmented Structures
           Segmentation                                         5
                                                                       BWH
                                        Large Scale
                                                                    Visualization
                                         Distributed
                                        Computing


 1
      Data Donor                         BIRN             Preliminary Study:
     Site (WashU)                       Data Grid              Scientific Goal:
                                                          •46 hippocampus data sets
       N=45         De-identification                     classify CPU hours, 4 TB
                                                          •30,000 patient status from
                      And upload        2                    morphometric results
                                                          data
SASHA: Shape Analysis Pipeline Results

                                                           Shape-derived
                     6 semantic dementia subjects
                                                           metrics can be
                                                           used to detect
                                                           class-specific
                                                            information


   21 control subjects             18 Alzheimer subjects
The BIRN Collaboratory Today
 Enabling collaborative research at 28 research
  institutions comprised of 37 research groups.
How does Rocks make it do that?
 Installs operating system software
 Turns individual servers into a “Grid”
   •   Portals & web servers
   •   Data grid for access & management
   •   Compute clusters
   •   Database servers
 Distributes, installs and updates 3rd party, domain-
  specific scientific software packages
 Updates system software
What would be better?
   Add/Improve security & performance monitoring
   Detect and capture configuration changes
   Track versions
   Ideally, reduce, reuse, recycle…
Desired System Qualities
             Basic System Software         Agile
              Operating System              • Fast response with updates
BIRN/Rocks    Security…                     • Self-help for developers
                                           Repeatable
 Software    Server Definition Software
                                             • Tracking of versions
              Apache/Tomcat
  Stack                                      • Tracking of deployments
              Globus…
                                           Modular/Flexible
             Application Software            • Handles unique site
                                               requirements
              Gridsphere
                                             • Handles unique project
              HID
                                               requirements
              Mediator
                                           Customizable
              Scientific Applications…
                                           Scalable
                                             • Highly automated
                     Custom                  • Supports addition of several
                                               sites per year, plus additional
                BIRN Server                    projects over 5 years
What’s involved in a single grid?
                    Rocks Central & YUM            A BIRN Grid
  14 Rocks Rolls    Rolls:
                     * RHEL4
    (-2 for OS)      * area51
                     * base
                     * birn               Portal/Web       BIRN Rack
                     * birnafs

  17 Custom Rolls
                     * birncondor                            GPOP
                     * birnportal
                     * birnsrb
                                             HID
                     * CentOS
                                                            GComp
                     * condor
                     * cvsserver
                                           Mediator         Nettools
                     * freesurfer
                     * gama-naregi 1.0
  CVS, SVN &         * gama-naregi 4.1                           NAS
                     * ganglia
  SRB Repos          * grid
                     * gridsphere         DB Server
                     * hardwareutils
                     * hid                Registry DB
                     * hpc
                     * java
                                                         GAMA Server
                     * kernel
                                            UMLS
                     * mediator
                                                            MyProxy
                     * nagios               HID DB
    Testbeds         * oracle                                Globus
                     * postgres
                     * sciapps             MCAT                  CAS
                     * sge
                     * srb34
                     * tomcat
                     * updates-CentOS
                     * webserver
Software Development & Integration


        Testbeds                   BIRN-CC
    - update local CVS/SVN   - large source into SRB
    - update tarballs        - updates RPMs
                                 - Makefile
    - update RPMs                - version.mk
                                 - *.spec.in
    - new config/install
                             - updates XML (rare)




                                  Rocks/YUM
                                    Server
Software Deployment


   Development Area                   Staging Area
                                                                  Production Area
  -Integrate software for      -Verify interoperation of latest
                               code                               - Stable
  many diverse sources
                               -Support demonstration of          - Reliable
  -Version control at system
  and sub-component levels     latest development efforts         - Facilitates research
                               without disruption to production
  -Rolling baseline
                               -Functional system/Beta
  -Integration and             Testing
  Functional Testing



   Rocks Development                   Rocks Staging              Rocks Production
         Server                           Server                      Server



   CVS                 SRB
What we love?
 Repeatability
   • All the web servers are the same
   • All the database servers are the same
 Flexibility
   • Mix & Match rolls
 Level of automation
   • Experienced person can “kick” a server in 5 minutes
   • IPs, hostnames, software configuration done
 Open Source
 Result: Not one grid - Many! Not one project - Many!
What drives us nuts?
                   Turn around time for updates
                   Steep learning curve
                   RPM building not standard
                   Build time large
                   Software developers are not co-
                    located with integrators
                   Reinstalling to get updates is
                    not an option
                   Lack of advanced roll
                    development training
More info?
 See the project website: http://www.nbirn.net
 Email vrowley_at_ucsd_dot_edu

								
To top