Hi Tomohiro- Yes, since this observer is a minimal piece of code for these demos, we decided that the ultimate RA client would request a circuit with a specific GLobal ID, and the Observer would be manually configured (or hard coded) to query on that Global Id. We will use a Query() to discover the path of the connection, then periodic Query()s to poll for state change. So...The observer will need to know which NSA to poll for each connection. We decided that the network name could be embedded in the GlobalID (I think this is already a best practice with IDCP). Else, we have a single client uRA that both requests the connection, and then subsequently acts as the observer which would make the well-known global name unnecessary. So the observer needs to map the network name to the NSA responsible and then to the soap endpoint... So the observer can easily do this if it has the same topology file the NSAs use. If we integrate the observer into the client uRA, then this netwrok-NS mapping is moot as we already know which NSA to query(). However, the location of the segment endpoints needs to be sent by observer to the Automated Earth. For FIA/SC I am adding "location" object to the topology associated with each network. The observer/uRA client will need the topology to obtain this location information associated with each network hop. BR Jerry On 10/7/11 3:00 AM, Tomohiro Kudoh wrote:
Hi Jerone, Henrik and all,
The query behavior is specified as:
* Supports querying reservations based on connectionId or * globalReservationId. Filter items specified are OR'ed to build * the match criteria. If no criteria is specified then all * reservations associated with the requesting NSA are returned.
Therefore, for an external agent such as viewer to get information by query, the external agent should know connectionIDs or globalReservationIds of reservations somehow.
Tomohiro
On Thu, 6 Oct 2011 13:20:16 +0200 Jeroen van der Ham<vdham@uva.nl> wrote:
On 6 Oct 2011, at 13:06, Henrik Thostrup Jensen wrote:
Hi
On Wed, 5 Oct 2011, Tomohiro Kudoh wrote:
To realize this, each ultimate provider should push state transition information to the google app engine. If everybody agree, we will prepare a sample code which should be incorporated to the implementations.
How do you think? Why not have the visualization tool use the NSI query interface? After all, it is what it is supposed to be used for. I dislike the idea of starting to build a parallel infrastructure for a task which is supposed to be solvable with the NSI protocol. Sure, the polling nature of query and the high entrance bar for SOAP+WSDL is probably not ideal, but I think we should at least try to build something using it, before we start on the parallel infrastructure. If it doesn't work or becomes to convulated or complex, perhaps we should reconsider parts of the protocol. If we cannot use NSI, who can? If the NSI query interface is implemented and we can easily get the required information out of it, then I'm all for using that.
In fact, I think that with your OpenNSA implementation I/we can jumpstart that implementation and should be able to query NSAs in no time at all.
Jeroen. _______________________________________________ nsi-wg mailing list nsi-wg@ogf.org http://www.ogf.org/mailman/listinfo/nsi-wg