NML 2009-03-03, morning session (Notes by Freek Dijkstra, Lars Fischer and Richard Hughes-Jones) WELCOME ======= Jeroen van der Ham and Martin Swany welcome everybody Jeroen is interim chair for this meeting along with Martin. Paola Grosso steps down as co-chair, Freek Dijkstra will follow her up after this meeting. DELIVERABLE 1 ============= deliverable sent out on mailing list no comments from group on deliverable 1. Deliverable #1 was sent to the area directors (Franco Travostino and Richard Hughes-Jones). Some doubt on procedures. Franco did ask for names of external reviewers. This was not required, as this is an informational document only (not a proposed standard). If the external reviewers have been asked, we will wait for their input. Otherwise, Franco and Richard will give informal feedback, and the document can be sent to the document editor. VXDL (PASCALE VICAT-BLANC PRIMET) ================================= Talk: Pascale Primet - VXDL: virtual exchange infrastructure description language. Slides: http://forge.gridforum.org/sf/go/doc15522 CARRIOCAS Project: service delivery over ultrahigh capacity networks - with a view to business models. The project describes short-lived overlay infrastructures over a fixed infrastructure. The language VXDL was developed to describes short-lived overlay infrastructures over a fixed infrastructure. It includes (1) list of individual resources and groups, (2) network topology, and (3) executing timeline. The topology description was developed in parallel to the NML, and it not equal, but the purpose is the same. Links can be described recursively (a link can contain other links). A UML diagram of VXDL is available at http://perso.ens-lyon.fr/guilherme.koslovski/download/vxdl.jpg Comparison of VXDL and NML elements (see slides) Pascale proposed to integrate VXDL with GLUE 2.0, which describes computing resources. Inder wondered how this works relates to JSDL, a job description language. Pascale: JSDL is for job description by end users, VXDL is used to describe virtual infrastructure by intermediate service providers. Suggestion to look more into this. Cees: are you describing all existing or available resources? Pascale: CARRIOCAS publishes functionality, not the availability. That is checked during reservation request. Comment: mapping between user request and and network provider descriptions. How do we map what VXDL (or JSDL) does into "what the network can provide" (as done in NML) - user requirement view vs. reseource / network capabilities view, and the mapping between those; answering "who can give me what I am asking for" Comment: there's a need for this - but this may not be for this workgroup; it needs to focus now and allow for this as a future extension. Pascale: virtualization is future Internet research. It is not clear were standardization should be done. IETF, IRTF, OGF, ... Inder: these communities (network and computing communities) do no understand each other at all, no common language. GLUE defines end-points, but no network. Martin: we try to form a liaison with them. For this reason, we asked a chair of GLUE for external review of the usage document (deliverable #1). Inder: other user still don't see the network as a configurable resource, and a collaboration is also an education to others. Short discussion on GENI and FIRE. Those communities currently are application-centric and lack repeatability of experiments. Virtualization working group had a notion of networks, but has been closed after inactivity. IDENTIFIERS =========== Summary talk: Jeroen van der Ham and Freek Dijkstra Slides: http://forge.gridforum.org/sf/go/doc15525 Identifiers contain a domain and opaque Parts. The structure of the opaque part is defined by each domain, and should not be interpreted by other domains. String-comparisons should be possible Several schemes are out there: NURN,URN, URL, GLIF. They are all basically the same We really need to make a decision - it's not very important which one we pick Agreement: we will propose "urn:ogf:network:<example.net>:[opaque part]" to the mailing list. URN:OGF: has to be registered. Richard will sent an e-mail to Joel Replogle. a. OGF will need to register "urn:ogf" with IANA (This requires IETF consensus action; per RFC3406, it seems to require a short informational RFC. Also it will need an OGF web page with the registration of the sub-divisions) b. OGF will have to approve use of the sub-division :network: