OGSA teleconference - 9 August 2004 ==================================== * Participants Peter Ziu (Northrop Grumman) Andreas Westerinen (Cisco) Jeffrin Von Reich (HP) Jem Treadwell (HP) Latha Srinivasan (HP) Andreas Savva (Fujitsu) Susan Malaika (IBM) Fred Maciel (Hitachi) Heather Kreger (IBM) Hiro Kishimoto (Fujitsu) Michael Behrens (R2DD, LLC) Minutes: Andreas Savva * WS Distributed Management (WSDM) tutorial - Presentation by Heather Kreger "OASIS Web Services Distributed Management Technical Committee Specification Overviews" http://tinyurl.com/5fxfd/WSDM-overview/en/1 - Other related WSDM material: - Fred's GGF11 slides (Introduction to WSDM) http://tinyurl.com/4zlfe/Introduction_to_WSDM/en/1 - [GGF11 CMM WSDM recording] http://tinyurl.com/4zlfe/GGF11_Session_1_MP3/en/1 - Initially chartered as the "Management Protocol TC" and then expanded scope and rechartered. - Many companies are represented (list on slide 3 is not exhaustive) - Management using WS (MuWS) - Historically management was one of the first areas that moved to distributed computing - Moving to describe and provide access to manageability using WS - MuWS: Uses web services as management platform - Management of WS (MoWS) - Web services are also IT resources to be managed - WSDM defines manageability of resource; not what managers do to manage resource. - Define basic features as they relate to management (properties, relationships, notifications, etc.). One example is identity. - It's not expected that people should re-implement to support WSDM; re-wrap through a representation to provide access to the 'real' resource. - [Slide 7] Two patterns shown: Resource ('Any real resource' provides manageabilitiy directly or resource is wrapped ('Resource representation'). In the past the Agent was also visible as a separate entity. Difference here is that the representation appears the same as the 'Real resource.' - WSDM removes the explicit need for agent; the agent may be there but it is transparent to the manager. Manager only 'sees' the resource. - This is one main difference when contrasted with past efforts - [Q] Relation to UDDI? - [A] WSDM will define some way to discover resources; not sure that UDDI is a good choice for such a directory however. - [Q] Does 'Resource representation' here refer to modeling using WSRF (e.g., is it a WS-Resource)? - [A] No. Resource representation just indicates that there is a proxy to the real IT resource. - Manageability defined using WSDL - Capabilities of resource - Lifecycle of the resource (operational states) - Relationships are capabilities too - And resource representations - Common web service porttypes - Lifecycle of the web service itself; not a comprehensive model - [Q] How detailed is the model description? - [A] WSDM is model agnostic - WSDM defines state change events, and some common basic state model. When an actual resource is represented it will have to extend the basic definitions to expose all its information. E.g., Adding new attributes or refining existing ones. - E.g., WSDM requires a state property; it is defined as a complex type but only very granular states are pre-defined. - About events: It is not required to emit them but if a resource does then those events must adhere to a certain format. - Existing models provide the content - Various models were analyzed and the common/core aspects were extracted - MUWS defines a very basic operational state diagram. As already stated resources are not restricted to this; it can be extended or something else can be defined instead. - OGSA also expects to define some state diagrams (e.g., job). Not sure how good a fit exists between OGSA and WSDM. - Agreed that this is one topic that is worth adding to the cross-wg agenda (>Hiro) - WSDM does manageability porttypes and resource representation as WS - In WSDM 1.0 there will be some definitions of relationships but not clear that these definitions should be part of WSDM - Provide end to end access to manageability (slide 12) - Low entry point: don't need specialized software; can use existing http/web infrastructure. - [Q] Some referenced specs are not standards or even have not been submitted to standards bodies (e.g., WS-Addressing) - [A] Dependencies on stuff that has been published. - Approach is that when WSDM 1.0 is released all referenced specs must have been (at least) submitted to some standards bodies - And there is commitment that WSDM 2.0 will use standardized versions of these specs. - Hoping that initial structure to handle meta data will be provided by WSRF - Event format - (Hiro) Grid would also need some common event format; and we have to discuss whether requirements match. - Agreed that this is another topic for the cross-wg session (>Hiro) - Policy and Negotiation (less sure about negotiation) - Future work. Probably not addressed until 2.0 or later - [Q] Relationship with WS-Agreement? - [A] Still not sure how negotiation applies to WSDM - Initially thought to allow for negotiation of policy (but is somewhat unclear at the moment) - Policy - Might need to define policy as applied to management - WS-Policy is somewhat misnamed; does not define policy. Rather it is focused on exchanging capabilities; how to do attachment of policy documents (so might be clearer to call it ws-attachment or something like that). - Policy should have data model: what action to do based on what is received; it is not just the exchange of information. - Definition of manageable resource - And capabilities - Implied contract: suppose that operational state is defined then can only be in one of those states. - Metrics - Not definition of a particular metric but information about metrics (what kind of metrics and how they can be used) - Metric is represented as a property but with different metadata - Capabilities - operational state (slide 19) - Definition of basic model - 'Degraded' is the wrong term since it associated a 'bad' meaning to this state. 'intermediate' or 'starting' may be more appropriate names for it. - [Q] Is there definition of an initial state? - [A] It is the state that the resource is discovered in; implies that the resource should be able to report back its state. - CIM reports 'unknown' until the state can be determined. - There hasn't been sufficient discussion on the substates yet. - MoWS - Based on W3C task force for states, metrics, endpoint - Capabilities of WS (slide 22) - Left side entries (with no namespace prefix) are MoWS - [Q] What is difference between MUWS and WSDM namespaces? - [A] WSDM is more low level (but there is no big difference; could make them the same without problem) - WSDM Spec 1.0 roadmap - [Q] Will an Event model be defined? - [A] WSDM defines event format/content. WS-Notification defines the notification model. - Extend the operational state: probably refining the state diagram - Handling discovery should also be on the list. - WSDM 2.0 - [Q] What does provisioning mean here? - [A] Create new instances of resources, or drive installation of resources. ** Other discussion - How to make sure that OGSA and WSDM taxonomies fit? - Start with comparison of what OGSA expects and what WSDM defines. Expect that it is one part of collaboration between the two groups. - Similarity with Java management extensions (JMX) - WSDM is language/model neutral - Distributed aspects: would there be discussion on the distribution and persistence of management data - Agreed that manageability can be provided by one or more web services; a resource's configuration data could be provided by some other resource. - Haven't decided if this is transparent or visible to the manager - E.g., naming: Might want to have unique identifiers that are unique across domains. - To some extent uniqueness is addressed (identity comparison). But different identities do not preclude that it is the same resource. - No rules to define unique identities - Work on data management might be relevant here Action: Hiro to write up agenda for cross-wg session using some of the topics identified in the minutes and post to both working groups in advance. * August F2F meeting discussion - https://forge.gridforum.org/projects/ogsa-wg/document/Agenda_for_2004_August_interim_meeting/en/3 - (or http://tinyurl.com/5w4eh ) - Will use 2 facilities: Intel and Fujitsu (FLA) - Intel meeting room has internet access but Ravi can't attend on Thursday so FLA facility will be used. - At the WSRF F2F there was an one hour summary session at the end of each day to let people who might want to dial in get a summary and also to focus the group on the agenda. - Given the time zone it might be better to put this review session in the morning; but there might not be enough time on Friday morning (only half day session. Action: Hiro to work summary sessions in the F2F agenda. - Hiro will not be attend on Aug 20 (schedule conflict). Ian will chair the meeting instead. * AOB - Jem, Latha and Fred will probably not be able to attend the Thursday session.