- From: Heather Kreger <kreger@us.ibm.com>
- Date: Thu, 19 Dec 2002 17:40:11 -0500
- To: www-ws-arch@w3.org
W3c MTF 12/18/02 Attendees: Igor - CA Heather - IBM Mark – Talking Blocks Yin-Leng - HP David Booth – W3C Discussion: Web Service Management Requirements Discussion: Service vs Port: are we managing WSDL.Service? or WSDL.Port? Service Instance or what’s exposed to consumer that we can measure against. Is defined as Service, or group of portTypes, at least one portType. Don’t go to ports because either do related interfaces (portTypes) or one portTypes with many ports (bindings) associated with it. Should manage what consumer and provider agree on portTypes… not the implementation that is exposed in port. Service itself is a single portType or group of portTypes. We’ll manage at levels defined in WSDL: by service, then by port and portType The following information should be available for each service as defined by its name in the WSDL: Information required to be manageable, where it comes from, format of it is to be decided by the development a detailed specification. We have discussed the following fields: Identification Identification information: o service identifier, required – uri – a unique identifier for the service instance or is it type?Refer to wsdl for specifics. Could be different syntax/multiple uri’s to refer to a service. Can deref to exactly the same description of the service. o service name, not required – readable name for the service, Does it need to match the service name or the port name in the wsdl? Can get name from WSDL document for service. Not required, but VERY convenient. Do we need to standardize optional fields? If provided should be the same as the value in WSDL. o service description, not required – readable description of the purpose of the service, this could be a catcher for taxonomy, Can get from WSDL, but its not required there either. If is provided should be same as in WSDL o service version, required – the version of the service as defined in the WSDL document, this information should ALSO be in WSDL, although at this time there is no standard way to define it. We will need to define a standard WSDL extension for this. § Vote: Mark, Heather, Igor: required; YinLeng: part of absolute minimum, mandatory o WSDL Reference, required – the reference for the wsdl document for the service, its somehow allows the manager to get a copy or access to the WSDL document. § Vote: Heather, Mark: Required; Igor: optional… not always at a URL, could be cached or genned on demand by infrastructure. Resolution, call it a reference, let followon work decide format.; Igor is now required. o Semantics URI, required – WSDL tns of a WSDL, identifies the semantics. Use identifier to refer to semantics or could use it to get a document that describes the semantics of the service, what you get depends on mime type. Could be a human readable doc, or rdf doc. Could be the URL of a document, but its not required to be a URL. § Vote: Igor, Mark, David, Heather?: required; YinLeng: optional; o (move to metric) Configuration Configuration information o WS Access URLs, optional – this would be the same is the URL in the location element in the port referring to this service. Can a service have several of these? For a service, tell all the access URLs. Move discussion to service instance. o Ws WSDL description URLs + service name + port name, already part of Identification. o Configuration files, remove – optional – filename that contains any configuration for this service. It should be an XML file. If there are configuration files they should be identified?Config data is really about a service instance and not for the ‘WSDL.Service’concept. Move to service instance This is as far as the discussion progressed. The following is the rest of the section to be discussed on our next call on Jan 9th. Configuration Information o Admin URL – (optional) URL to access that supports administering this service o Admin WSDL URL o Management URL – URL to access that supports a w3c recommended management protocol to access the data defined in this specification for this service o Management WSDL URL – the url for the management url’s WSDL. If we have this, do we need management url since it’s the location in the port definition and may vary for read/only vs reconfigure ports. o o Security settings – optional This needs more thought and exploration with ws-security Metrics Metrics to help track usage of the service and execution environment. This was intended to be counts for the life of the service.. (since it was enabled, not since instantiated) or WSEE. Should we keep timestamps for each metric to scope the counts/averages? · Availability time – what time the service was enabled. · Number of Requests – total number of web services requests for this service · Number of Success Responses – total number of success responses sent by this service We need to define ‘successful’ · Number of Failure Responses – total number of failure responses sent by this service . We need to define ‘failure’ response · Average Response Time of Responses – the average response time for all responses from this service. The timestamp should be taken as soon as it enters the service and again just as it responds. It will be the aggregate of the time spent in the service itself. · Average Response Time of Failure Responses - the average response time for all failure responses this service. The timestamp should be taken as soon as it enters the service and again just as it sends the response. It will be the aggregate of the time spent in the service itself. · Average Response Time of Successful Responses - the average response time for all successful responses for this service. The timestamp should be taken as soon as it enters the service and again just as it sends the response. It will be the aggregate of the time spent in the service itself. · Total Elapsed Execution Time – the total ms spent processing in the web service for all requests (can then calculate average) · Success Response Execution Time · Failure Response Execution Time (or is this total-success) · Number of Invocations Per Method – the number of requests for each operation/method supported by the service · Number of Successes Per Method · Number of Failures Per Method - the number of failures returned by the service for each operation/method supported by the service (or is this total-success) · Average Response Time of Responses per Method - the average response time for all responses for each method. The timestamp should be taken as soon as it enters the method and again just as it exits the method. It will be the aggregate of the time spent in each method. · Average Response Time of Failure Responses per Method - the average response time for all failure responses for each method. The timestamp should be taken as soon as it enters the method and again just as it exits the method. It will be the aggregate of the time spent in each method. · Average Response Time of Successful Responses per Method - the average response time for all successful responses for each method. The timestamp should be taken as soon as it enters the method and again just as it exits the method. It will be the aggregate of the time spent in each method. · Success Elapsed Execution Time per Method · Failure Elapsed Execution Time per Method (or is this Total-success) · Total Elapsed Execution Time per Method – the total ms spent processing this method in the web service for all invocations · Number of Invocations per Attachment (?) – number of times an attachment is sent to the web service · Number of Invocations per Message (?) – number of times a message element is sent to the web service (for doc styles counting?) · High/Low water marks? · Rate of invocation? Operations Operations to control service lifecycle, deploy, remove, start, and stop are part of the execution environment’s operations. They could be moved to the service specific MBeans, but it’s a 50-50 choice so for these examples, they are part of the execution environment MBean. · Deploy – have service start its deployment · Undeploy – have service start its undeployment · Enable – allow this service to be invoked to respond to a request · Disable – stop this service from being invoked to respond to a request Events The notifications should be sent by the execution environment for each service, including: o service not deployed – tried to enable or request to a service that is not depoyed in wsee o service unavailable – tried to request on a service that is deployed an dis not enabled o service failed – request to the service failed o service deployed – lifecycle event that a service has been deployed service access denied – security denial o service invocation failed – a request to the service has failed within the service implementation (not via the wsee) o service not found –the service is deployed and enabled but not invocable o service timed out – request to the service timed out Where do we get the timeout value? o service deploy failed – deployment of the service failed o service enable failed – enabling the service failed Heather Kreger Web Services Lead Architect STSM, SWG Emerging Technology kreger@us.ibm.com 919-543-3211 (t/l 441) cell:919-496-9572
Received on Thursday, 19 December 2002 17:40:45 UTC