300 likes | 485 Views
CSF4 Meta-Scheduler Tutorial. 1st PRAGMA Institute Zhaohui Ding zhding@ucsd.edu or zhaohui.ding@email.jlu.edu.cn College of Computer Science & Technology Jilin University National Biomedical Computing Resource, University of California, San Diego. Agenda.
E N D
CSF4 Meta-Scheduler Tutorial 1st PRAGMA Institute Zhaohui Ding zhding@ucsd.edu or zhaohui.ding@email.jlu.edu.cn College of Computer Science & Technology Jilin University National Biomedical Computing Resource, University of California, San Diego
Agenda • Meta-scheduler & CSF4 Introduction • CSF4 Architecture • CSF4 Functionalities • Future Work • Demo and Practice
What is Meta-Scheduler • Resource Allocation & Management • Heterogeneous • Distributed • Dynamic • Local Scheduler VS Meta-scheduler
Meta-Scheduler VS Local Scheduler • Local Scheduler • LSF (Load Sharing Facility) • PBS (Portable Batch System) • SGE (Sun Grid Engine) • Condor • IBM Loadleveler • Meta-Scheduler • CSF • Maui (Silver) • Gridway • Nimrod-G • Condor-G
CSF4 • What is CSF Meta-Scheduler • Full Name: Community Scheduler Framework • CSF4 contains a group of grid services host in GT4 • CSF4 is a full WSRF compliant meta-scheduler. • Open Source project and can be accessed at http://sourceforge.net/projects/gcsf • Developed by Lab. of Distributed Computing and System Architecture, Jilin University, China • CSF4 has been added to Globus Toolkit 4 as an Execution Component
What CSF4 Can Do? • Basic Functionalities • Submit jobs to Grid without Specifying Cluster • Monitor and Control Jobs • Provide Queuing Service • Schedule jobs and resource by custom-built polices • CSF4 Portlet (A Web browser based User Interface)
What CSF4 Can Do? (cont.) • Advanced Functionalities • Multiple Domains Resource Information Sharing • Multiple scale resource scheduling policies • Automatic user credentials delegation • Automatic data-staging • Extensible scheduling framework • Supporting grid parallel jobs (MPI&MPICH-G2)
Local Scheduler And Infrastructure Supported by CSF4 • Local Scheduler Supported • LSF • PBS • SGE • Condor • Infrastructure Supported • Globus Toolkit 4 • Globus Toolkit 2
CSF4 – Functionalities Scheduling Plug-in Framework • Designed For Queuing Service • Provide A set of policies • Customizable • Extensible
Existent Scheduling Policies • FCFS (First Come First Serve) round-robin • Default policy • Throttle • Restrict the number of jobs in a scheduling cycle • Array Job Plug-in • Design for life science applications (such as AutoDock, BLAST) • MPICH-G2 Plug-in (under-developing) • The plug-in guarantee the synchronized resource allocation can be successful • Data intensive applications plug-in (under-developing)
Schedule plug-in & scheduling policies • Each policy is implemented inside a scheduling plugin module • A queue can load multiple plugin modules
Resource Information Sharing • A MDS information provider for CSF4 • Multiple CSF4 can share the resource information
CSF4 – Functionalities (cont.) • Deploy Multiple CSF4 in a Grid Community
Array Job • AutoDock and Blast-like applications • A large number of sub-jobs. • Execute same binary • Different input/output files
Array Job (cont.) • Advantages • Submit job only once • Save submission time and memory storage
Data Staging • Manual Data Staging • Which clusters I can use? • Which clusters my jobs will running on? • Where is the output data? • When will the job finish, so that I can stage-out the output data?
Manual Data Staging • Without Meta-Scheduler
Automatic Data Staging • With CSF4 Automatic Data Staging
Integrate CSF4 with Gfarm • With CSF4 Automatic Data-Staging and Gfarm
CSF4 User Interface • CSF4 Portal
CSF4 User Interface • CSF4 Command Line
Demo & Practice • https://www.nbcr.net/pub/wiki/index.php?title=CSF4_Tutorial_PRAGMA13
Thank you감사합니다ありがとうございます謝謝谢谢