Скачать презентацию Portal Software Browser-based Monte Carlo Zisis Papandreou University Скачать презентацию Portal Software Browser-based Monte Carlo Zisis Papandreou University

8db73947f0e4ca0cf25a7ccc668498ee.ppt

  • Количество слайдов: 16

Portal Software: Browser-based Monte Carlo Zisis Papandreou University of Regina Glue. X Collaboration Meeting Portal Software: Browser-based Monte Carlo Zisis Papandreou University of Regina Glue. X Collaboration Meeting JLab, December 11 -13, 2003 SPARRO Group, University of Regina 1

Outline n Grid 2003 Conference q n Grid Services Regina Portal Application framework, Globus Outline n Grid 2003 Conference q n Grid Services Regina Portal Application framework, Globus q Meta queuing q n Collaboration tools: q Glue. X Forum, Database Administrator SPARRO Group, University of Regina 2

Grid 2003 Conference (Ed, Zisis) n n n Phoenix, AZ, November 16 -17, 2003 Grid 2003 Conference (Ed, Zisis) n n n Phoenix, AZ, November 16 -17, 2003 Concurrent with SC 2003 (8, 000 hackers) Interactive and coached tutorial: n How to Build a Grid Service using the Globus Toolkit® 3 (Argonne people) n n Written in Java Built (JDK) and run on Ed’s RH 9. 0 linux laptop SPARRO Group, University of Regina 3

OGSA GT 3 Architecture n n n SPARRO Group, University of Regina Core: Grid OGSA GT 3 Architecture n n n SPARRO Group, University of Regina Core: Grid Services Security: SSL, X. 509 Base: Job Services, Index Services, RFT Data: Replica Management (data subsets) Other: foreign grid services 4

Creating a Grid Service Web Services Description Language n n n Write WSDL directly Creating a Grid Service Web Services Description Language n n n Write WSDL directly Generate WSDL from interface language (Java) Decorate it Generate stubs Implement it package gt 3 tutorial. core. first. impl; public interface Math { public int add(int a, int b); public int subtract(int a, int b); public int multiply(int a, int b); public float divide(int a, int b); } javac Math. java org. apache. axis. wsdl. Java 2 WSDL -P Math. Port. Type -S Math. Service -l http: //localhost/ogsa/services/core/first/Math. Service -n http: //first. core. gt 3 tutorial/Math gt 3 tutorial. core. first. impl. Math SPARRO Group, University of Regina 5

Exercises accomplished q q q Grid Services / GT 3 File. Share Service deployment Exercises accomplished q q q Grid Services / GT 3 File. Share Service deployment File sharing with identifiers Adding service data Register with community index SPARRO Group, University of Regina q q q Lifetime management File searching Security Transience: create and destroy File. Shares Grid Service Publishing 6

Demonstration: File. Share Grid Service (Ed Brash) SPARRO Group, University of Regina 7 Demonstration: File. Share Grid Service (Ed Brash) SPARRO Group, University of Regina 7

Regina Portal: Beefed up https: //www. phys. uregina. ca/Glue. X/ n n Glasgow: job Regina Portal: Beefed up https: //www. phys. uregina. ca/Glue. X/ n n Glasgow: job submission, review (PHP) New features: Cold. Fusion Application Framework (CFML) q User registration & authentication q Integration with Grid. Port TK, Globus q Upload XML Geometry and compile q Experiment or commit executables to db q SPARRO Group, University of Regina 8

Demonstration: Regina Portal (Zisis Papandreou) SPARRO Group, University of Regina 9 Demonstration: Regina Portal (Zisis Papandreou) SPARRO Group, University of Regina 9

Grid Computing q q “The Grid”: distributed computing infrastructure (both software and hardware) that Grid Computing q q “The Grid”: distributed computing infrastructure (both software and hardware) that provides access to high-end computational capabilities Grid computing enables the shared use of computational resources such as data archives and simulations. Resource owners offer resources subject to constraints Resource consumers also place constraints on properties SPARRO Group, University of Regina 10

Queuing, Scheduling & Meta Scheduling (Aadila Soobrattee) n n n Regina open. Space cluster Queuing, Scheduling & Meta Scheduling (Aadila Soobrattee) n n n Regina open. Space cluster (50 CPUs) is governed by Open. PBS local queuing software Need capabilities of scheduling workload among a computer grid where local scheduling software and local policies may differ: meta-scheduler Schemes: n n n Centralized: surface info, locals dispatch jobs Hierarchical: meta and local share info; no flexible re-routing Distributed: scheduler/dispatcher model; flexible, scalable SPARRO Group, University of Regina 11

Distributed Scheduling n n Meta schedulers at every site Jobs submitted to meta scheduler Distributed Scheduling n n Meta schedulers at every site Jobs submitted to meta scheduler where job originates Meta schedulers query each other to collect instantaneous load info; job is transferred to the site with the lowest load Dagman, Condor/G, Nimrod/G, GT 3/Platform, Silver/Maui SPARRO Group, University of Regina 12

Silver/Maui Silver: • reduces average job queue time by enabling access to a larger Silver/Maui Silver: • reduces average job queue time by enabling access to a larger set of computer resources • intelligently decomposes jobs into multi-system sub-jobs to improve job turnaround time • manages site-dependent job environment and resource requirements • supports jobs which require co -allocated resources which expand multiple systems • supports jobs too large to run at any single site by spanning jobs across multiple sites ü Silver provides grid scheduling services for clusters using Open. PBS, can operate on all major UNIX operating systems and is capable of operating with or without Globus http: //www. supercluster. org/ SPARRO Group, University of Regina 13

1. User logs into Globus and gets Globus credential 2. User submits job from 1. User logs into Globus and gets Globus credential 2. User submits job from site A via sqsub client 3. sqsub translates job to global job and stores job and credential in SILVER's global queue 4. SILVER determines list of feasible resources based on resources requested 5. SILVER queries Globus and determines machines that support the specific credential 6. SILVER queries each feasible resource and obtains matrix of resource availability times 7. SILVER determines resource associated with earliest job completion time (site B) 8. SILVER reserves resources on site B 9. SILVER stages input data and executable to site B using Globus file transfer (gssftp) 10. SILVER translates global job to RSL and submits job to Globus with 'run on site B' request 11. Globus translates RSL job to resource manager command language and submits the job to resource manager on site B. 12. MAUI detects job, executes it on reserved resources 13. job completes 14. MAUI creates accounting record, removes reservation, and updates SILVER 15. Output data is staged to requested destination using Globus file transfer SPARRO Group, University of Regina 14

Collaboration Tools n Glue. X Forum (at UConn) n http: //zeus. phys. uconn. edu/forums/gluex. Collaboration Tools n Glue. X Forum (at UConn) n http: //zeus. phys. uconn. edu/forums/gluex. org/index. php n Web-based email forum (written in PHP) Category-based email, threads, moderators n n Postgre. SQL Database Administration n pg. Admin. III (win & unix) Remote connection to db servers Create/delete tables, enter data, view data SPARRO Group, University of Regina 15

What’s next? n Expand portal further: n n Install Meta Queuing System at Regina What’s next? n Expand portal further: n n Install Meta Queuing System at Regina Web/Grid Services (document uploading, indexing and searching) XML Data Transfer (Web Distributed Data Exchange – WDDX) Communications: n n n Teleconferencing: AT&T Videoconferencing: VRVS or Polycom Email threaded discussions: Glue. X Forum SPARRO Group, University of Regina 16