SURAgrid Accounting Status September 2007

16
SURAgrid Accounting Status September 2007

description

SURAgrid Accounting Status September 2007. SURAgrid Accounting Status September 2007. Accounting WG - Phase 1. Call for a working group at SURAgrid All-Hands March 07, to develop (at least) a recommendation for implementing accounting on SURAgrid - PowerPoint PPT Presentation

Transcript of SURAgrid Accounting Status September 2007

Page 1: SURAgrid Accounting Status September 2007

SURAgrid Accounting StatusSeptember 2007

Page 2: SURAgrid Accounting Status September 2007

SURAgrid Accounting StatusSeptember 2007

Page 3: SURAgrid Accounting Status September 2007

Accounting WG - Phase 1

• Call for a working group at SURAgrid All-Hands March 07, to develop (at least) a recommendation for implementing accounting on SURAgrid

• Group formed with desire to be short-term but contribute a tangible immediately useful first step – Volunteers included: Victor Bolet, Steve Brandt, Mahantesh

Halappanvar, Steve Johnson, Shawn McKee, Jerry Perez, John-Paul Robinson, Bin Zhou

Page 4: SURAgrid Accounting Status September 2007

Accounting WG Phase 1

• Goals of initial SURAgrid accounting:– To develop and implement the capability to analyze

SURAgrid usage in support of setting implementation priorities, planning, and input to proposals.

– To monitor allocations, however these are prescribed by the resource owner at this time.

– Create a system that can be implemented with current resources but also incorporated into, or expanded by, more complex system later on if needed - also tracking well with standards and best practices in this area.

Page 5: SURAgrid Accounting Status September 2007

Phase 1 Implementation Plan

• Determine set of common data to be collected from SURAgrid resources

• Document desired data collection in the form of a SURAgrid Usage Record– Based on job Usage Record standard currently

progressing through the OGF User Record Working Group (http://forge.ggf.org/sf/projects/ur-wg).

• Enable data collection through local schedulers• Store data in standard format (XML) and central database

for access now, also later availability to another accounting package if one is selected.

Page 6: SURAgrid Accounting Status September 2007

Snapshot Status Phase 1 Work

• Some still to discuss but…SURAgrid Usage Record essentially complete!

• WG members piloting with all major local schedulers & documenting for others to follow

• Data from pilots is being stored in centralized database w/querying capability, hosted by ODU

• Will need to see if initial database & reporting package provide enough utility, can scale, etc. – “Accounting WG Phase 2” to address if not…

Page 7: SURAgrid Accounting Status September 2007

With Special Thanks to…

This was a real working working group and extra appreciation goes to the following SURAgridr’s for producing the components you see today:– Victor Bolet– Steve Brandt– Mahantesh Halappanvar– Steve Johnson– Shawn McKee – John-Paul Robinson

Page 8: SURAgrid Accounting Status September 2007

SURAgrid Usage Record

See handout…

Page 9: SURAgrid Accounting Status September 2007

Status of Local Scheduler Pilots

• PBS: Conversion to XML complete, script provided, incorporated into workflow

• LoadLeveler: Conversion to XML complete, script provided

• SGE: Basic data mapping defined, needs conversion to xml

• Condor: Conversion to XML complete, script provided, OSG Gratia project

• GridWay: Initial investigation suggests data available, needs code to extra from data store

• LSF?

Page 10: SURAgrid Accounting Status September 2007

Working Group Action Items

• GSU: work on Loadlever scripts with LSU

• ODU, TAMU, LSU: Loadlever and pbspro upload workflow

• TAMU: hammer out PBS to URWG conversion and will look at user mapping

• ODU, UAB, TAMI: will working on running 1.4.3 xindice (parallel development path with work-flow effort)

• UAB: SGE and Gridway acct data record generation, project page updates, move wiki to more formal home to facilitate future collaboration

• ALL: solidify project documentation with ideally step-by-step instructions

Page 11: SURAgrid Accounting Status September 2007

Status of User Documentation

• Currently in a Wiki, hosted by UAB

– http://webapp.lab.ac.uab.edu/projects/jpr/wiki/SURAgridAccounting

– Moving to UABgrid collaboration environment: http://projects.uabgrid.uab.edu/sg-accounting

• Working Group members still adding detail to this resource

• Still need to determine best flow for information

• Eventually move to other location or format?

– Can be hosted by UABgrid indefinitely

Page 12: SURAgrid Accounting Status September 2007

From PBS to Central Data Store

• Resource Manager: OpenPBS/PBS-Pro/Torque

• XML Conversion Tool: PBS-XML Accounting Toolkit

• XML Database: Xindice

• Report Generator: DarkSlide

Figure 1. A Basic Flow Chart for Collecting SURAgrid Usage Records.

Page 13: SURAgrid Accounting Status September 2007

PBS to XML

• Site: install PBS-XML Accounting Toolkit• Requires Java (1.4) and Ant• Export code from CVS @ Sourceforge• Build and Install RPM

% ant buildrpm

% rpm –ivh pbsxmlaccounting.rpm

• Set up user environment• Run command

• Central: install Darkslide and Xindice

Page 14: SURAgrid Accounting Status September 2007

Status of Central Data Collection

• Update from Mahantesh

Page 15: SURAgrid Accounting Status September 2007

Additional considerations

• Need to revisit the “for further discussion” fields in the SURAgrid Usage Record

• Need a way to map local user to grid user (dn)• On account naming conventions: Thinking at this time is to

leave it to sites and work on mapping.• It's up to local system to filter out non-suragrid users. Sites

could leave all accounting data in place and use SURAgrid summary to generate local reports. Might be a good way to generate a community around this resource.

Page 16: SURAgrid Accounting Status September 2007

Next Steps

• Need longer term Accounting WG for Phase 2– Insure and support implementation of SURAgrid Usage

Record– Evolve record as necessary– Evaluate use and utility of the basic system and

consider/select other options if needed– Participants:

• From the community: Mahantesh & Steve B. so far committed. Other volunteers?

• From SURA: Dali Wang, MFY if/as needed