Management Area Session OGF19, Chapel Hill Ellen Stokes and Hiro Kishimoto, Management Area co-Directors 29 January 2007
AGENDA 6pm-6:30pm Area Workgroup View Workgroup Status 6:30pm-7:30pm Landscape and Roadmap (moving forward) Discussion, Q&A But first, reminder of Standards Area all-hands sessions Wed Jan 31 9am-12:30pm “Integrating the Work of OGF” “Very Brief Area Overviews” “What’s Next for Area Interactions?” 2
Status at a glance GFD-R publication Interoperability exp doc CDDLM-WG ACS-WG GFD-R.P publication UR-WG RUS-WG Public Comment Review Charter WG at OGF19 GLUE-WG 3
Four independent implementations CDDLM-WG (Configuration, Description, Deployment, Lifecycle Management WG) Document publication All 5 documents from charter are published The 5th document (XML-CDL, GFD.85) published Jan 2007 Four independent implementations HP, Softricity (now Microsoft), NEC, and UFCG Interoperability experimental document Three documents are underway “Testing a Specification” paper is about testing methodology based on CDDLM spec experience Topics Expecting new chair(s) New topics: TBD 4
UR-WG (Usage Record WG) Document publication Version 1 is in Author’s position; it will be going to the Editor shortly Describes a common format with which to exchange basic accounting and usage data over a grid There are two independent implementations Interoperability experimental document No plan yet Next steps Discussion has begun on Version 2 More object-oriented and accommodates lessons learned from early adopters More information and discussion slated for OGF20 in Manchester Workgroup is not meeting at OGF19 5
RUS-WG (Resource Usage Service WG) Status New charter with slightly new goals and new milestones developed Two specs: RUS “core” and on top of this the RUS “extended” spec. Both specs should co-exist, but the extended included more features RUS “core” specification will be slightly revised and finalized for public comment (Mar 2007 – planned submit to editor for public comment) Independent Implementations of the RUS interface GridSAM (used within the UK e-Science infrastructures) UNICORE (used within the DEISA infrastructure) SweGrid Accounting System (SGAS) (component of Globus Toolkits) Distributed Grid Accounting System (DGAS) (used within the EGEE) Planned: LCG – RUS (used within OSG, GridPP) Interoperability work within the European OMII – Europe project Interoperability of RUS interfaces among UNICORE, SGAS, and DGAS Next Steps Collect experiences and feedback from early implementations Feature poll for a RUS “extended” specification on top of core RUS spec. Session: Wed Jan 31, 6pm-7:30pm (Bellflower) Discussions on the revision of the specification needed to support aggregate operations 6
ACS-WG (Application Contents Service WG) What is going on in ACS Defining a standard interface for the storing and exchanging the archives of the Application Contents. Repository Interface (ARI) + Archive Definition (AAD) Document Status Application Contents Service Specification 1.0 (GFD.73); September 4, 2006 Promoting the adoption further. Next steps Discussions underway (Enhanced interoperability , Repository Replication, Converged AAD…) Implementations Business Grid Project released sample implementation of ACS NAREGI project adopted ACS for its PSE part and its enhancement is underway. Session: Tues Jan 30, 4-5:30pm (Windflower) Introduction of NAREGI PSE implementation of ACS and Replication feature Revisit of the ACS 1.0 specification and its value Call for Adoption 7
GLUE-WG ** new workgroup** (GLUE Schema 2.0) Focus Recommendation document for abstract information model to facilitate interoperation between grid infrastructures Major revision of Glue 1.3 to address a few major problems and to incorporate more use cases Interoperability work done in the gin-info group has demonstrated the need for this work Other efforts within OGF will influence the work (e.g. JSDL, OGSA, BES) and used to fill gaps in groups that don’t have models of fundamental grid objects Multiple renderings Work will feed back to DMTF to incorporate Grid objects and their attributes into the CIM schema Use case input Glue schema 1.3 NorduGrid NAREGI CIM vendor extension Others as identified/evaluated Background (Version 1.3) Continues to be rolled out across production grids Rendered in LDAP, XML, and the Relational Model Used with OSG, Teragrid, EGEE, APAC and in different systems, MDS2, BDII, MDS4, R-GMA Chairs Laurence Field Balazs Konya Sergio Andreozzi Session: Thu Feb 1, 4pm-5:30pm (Azalea) 8
Landscape – The Basic Building Blocks (1) Management framework High level architecture / taxonomy that ties together pieces that comprise management (2) Management Services Logging, accounting, license mgmt, discovery, availability mgmt/metrics, performance mgmt, etc (3) Management interface / protocol for each resource e.g. Container, storage, network, job/services, security (4) Information / data model for each resource Key observation: Not all above work needs or should take place in management area Management area should address those items that are of wide applicability to all areas Those that don’t meet above assertion should be placed in other areas’ workgroups (e.g. best done in a specific area because of scope / expertise) 9
Management Framework (gaps in red) High level architecture / taxonomy Leverage existing documents, e.g. SCRM Landscape, Open Grid Services Architecture, Resource Management in OGSA, including use cases Use it to Identify the necessary pieces for management of grids Identify the gaps Drive new workgroups or revise/extend existing workgroup charters Identify, define, and implement specifications / profiles for a staged set of usable and interoperable management pieces to move adoption of grids forward in a very time efficient manner 10
Management Services (start of list) (gaps in red) Accounting service: The collection of resource consumption data for the purposes of capacity and trend analysis, cost allocation, auditing, and billing (out of scope). Accounting management requires that resource consumption be measured, rated, assigned, and communicated between appropriate parties (IETF 2975). Metering service - defines interfaces for collection and processing of metrics data for resource consumption (usage and duration) and aggregating results of those measurements by workload, application, job, or user Rating service - Rating translates technical measurements into monetary units. It is the process of calculating the cost and/or price of a resource by using its rating package (defining for example price per consumption) and a given metering record that measures the consumption of that resource. Logging service – keeps log records in a persistent store for a period of time, acts as an intermediary between log artifact producers and consumers, filters for records License management service – manage licenses in grid environment Discovery service, e.g. service and resource – is OASIS/WSDM (MUWS – Management Using Web Services) sufficient? Performance management service - monitors performance of resources, reports if metrics are out of expected ranges and values, may trigger operations to alter the resource’s configuration or the resource’s environment’s configuration in response to performance trends. (Infrastructure: nm-wg?) Availability service / metrics? Relationship to CDDLM? Usage Record (Management:ur-wg) Resource Usage Service (Management:rus-wg) Application Content Services (Management:acs-wg) Configuration, Description, Deployment, Lifecycle Management (Management:cddlm-wg) Others – existing or gaps? 11
Management Interface/Protocol (start of list) (gaps in red) Event format – relationship to OASIS WSDM, DMTF indications, needed for eventing protocol content; no WS-MAN alternative GFS-I.025: Analysis of ‘Top N’ Event Descriptions GFD-R-P.022: Distributed Resource Management API (Applications:drmaa-wg) Storage Resource Manager interface (Data:gsm-wg) Others – existing or gaps? Job state/status (drmaa, jsdl, bes, ur, event,…?) need consistency or how existing states/statuses relate 12
Information / Data Model (start of list) (gaps in red) Information model direction (Arch:OGSA-WG) Fundamental grid objects (Mgmt:glue-wg, Compute:ogsa-bes-wg, Compute:jsdl-wg) Service modeling Data modeling Others – existing or gaps? 13
Where do we go from here? Start investigation special interest group to lay out roadmap/topics, identify gaps, etc? Management requirements Leverage/liaise Enterprise Grid Requirements research group work/thoughts? Synergies with other OGF areas / workgroups? Evolve existing workgroups in the management area? Start new work groups? In management area? In other areas? What work can we leverage from other SDOs? Does it need modification or extension for grid? 14