[Nml-wg] Meeting notes NML session OGF 33

Tue Sep 27 05:27:01 CDT 2011

Attached are meeting notes of the NML sessions during OGF33.

Thanks to Jeroen van der Ham for keeping notes during the second
session. All errors in here are mine, please send corrects to the list.

The document is also available at
http://forge.gridforum.org/sf/docman/do/listDocuments/projects.nml-wg/docman.root.meeting_materials.ogf33

Meeting notes NML session OGF 33
================================

NML had three sessions, two on Tue 20 Sep, and one on Thu 22 Sep.

The first sessions consisted of formal presentation of related work:
* Jason Zurawski – Circuit Monitoring
* Jeroen van der Ham – GEYSERS; NOVI; GLIF DTOX-TF

The second sessions contained presentations with discussions:
* Jerry Sobieski – Jerry's Topology Wishlist
* Jeroen van der Ham – Port vs. Interface
* Freek Dijkstra – URN:OGF:network Status and Proposal
* Freek Dijkstra - Status and Work Items
* Jeroen van der Ham & Freek Dijkstra – Label Proposal

The third session was devoted to a technical discussion on the XML
syntax for labels, adaptations and switching services.
* Jason Zurawski - XML syntax

Jason Zurawski – Circuit Monitoring
-----------------------------------
Slides:
http://www.ogf.org/OGF33/materials/2350/20110920-CircuitMonitoring.ppt

Jason presented slides describing the circuit monitoring work by
Internet2, done late 2010, early 2011.
Some of these slides were presented earlier in the NMC working group,
but this time, Jason focussed a bit more on the schemas used.

Live demo: http://ndb1.internet2.edu/circuits/

Q: Why is there a topology collector (that queries both OSCARs and a
topology service)?
A: Topology service has static topology info, OSCARs has the dynamic
information.

One of the slides had a screenshot describing the following circuit:
-> ion.internet2.edu -> dragon.maxgigapop.net ->
<- dragon.maxgigapop.net <- ion.internet2.edu <-
This was confusing, as both dataflows seem to go from Internet2 to Max
Gigapop.

It was unclear if the bandwidth utilization of an interface lists the
total bandwidth of that interface, or just the bandwidth of a VLAN on
that interface. (The GUI does not list the VLAN number, so we presumed
it listed the bandwidth utilization of all VLANs on that interface).

The source code of the project is available online.
Code base topology service available at
http://anonsvn.internet2.edu/svn/perfSONAR-PS/trunk/perfSONAR_PS-TopologyService
Code base GUI is available at
http://anonsvn.internet2.edu/svn/perfSONAR-PS/branches/aaron-circuits/perfSONAR_PS-CircuitMonitoringAgent/

Jeroen van der Ham – GEYSERS; NOVI; GLIF DTOX-TF
------------------------------------------------
Slides:
http://www.ogf.org/OGF33/materials/2350/1109-OGF-NML-UvA-Activities.pdf

Jeroen presented three slide decks:
- on the GEYSERS project that the UvA is involved with
- on the NOVI project that the UvA is involved with
- on the Automated GOLE demo at GLIF, and the DTOX (dynamic topology
exchange) workgroup of the GLIF

Jeroen invited participants to join a BoF session at SC'11 on "Semantic
Resource Descriptions in Advanced Cyberinfrastructures". URL:
http://sc11.supercomputing.org/schedule/event_detail.php?evid=bof116

Both GEYSERS and NOVI use semantic (= OWL-based) topology descriptions.
Jeroen explained their schemata.

GEYSERS:
Richard asked why a "Network Element" is different from a "Node
Component". Would it be possible to integrate the network with other
resources (such as CPU, memory and storage)?

NOVI:
Slide manager - manages virtual pool of composed resources from
different domains
Aggregate manager - manage resource pool under single management
In a future plan, the ASB will be responsible for composition and
slicing of resources across domains. However, the Slice manager can not
be removed from the SFA, since not all (PlanetLab) domains have a ESB.

DTOX:
Discussion if NSI Topology represents the data plane or the control plane.
According to Jerry, the NSI Topology represents the data plane,
according to Jeroen the control plane.

According to Jeroen, an endpoint is combination of URN and label

Jerry Sobieski – Jerry's Topology Wishlist
------------------------------------------
Slides: http://www.ogf.org/OGF33/materials/2350/NSI+Topology+Wishlist.pptx

Jerry presented two slides with a wishlist that NSI has with regards to
topology description.

Jeroen van der Ham – Port vs. Interface
---------------------------------------

Jeroen argued that the change of the term "Port" to the term "Interface"
in NML would reduce confusion.
Both the NOVI and GEYSERS project use most of the concepts defined in
NML, and also use the same terminology, with the exception of the term
"Port", which is changed to "Interface" by both projects. The term
"Port" was deemed confusing with the ports in TCP and UDP.

Jeroen noted that a TCP port is actually a label on an Port/Interface.
There's no real advantage to changing the term other than to reduce
confusion.

A vote will be held on the mailing list to decide on the term in NML.

Freek Dijkstra – URN:OGF:network Status and Proposal
----------------------------------------------------
Slides: http://www.ogf.org/OGF33/materials/2350/NML+Progress+OGF33.pptx
(slides 8-11)

Freek presents progresson URN:OGF(:network)
- URN:OGF has been accepted by IETF, waiting for sister-doc in OGF
- document for registering subnamespaces is now in public-comments.
- urn:ogf:network request draft has been written

Problem with current urn:ogf:network syntax: domain names are somewhat
volatile.
Creating a registry for organisations that assign URNs would solve the
problem.
  - suggestion for the registry:
     - organisation name
     - DNS + version
     - DNS + year

Freek will write a proposal to the mailing list.

Jeroen van der Ham & Freek Dijkstra – Label Proposal
----------------------------------------------------
Slides: http://www.ogf.org/OGF33/materials/2350/NML+Progress+OGF33.pptx
(slides 12-13)

Jerry and Jeroen volunteer to provide a candidate solution for
expressing this in the UML.
IP and MAC addresses as labels?

Freek Dijkstra - Tasks Ahead
----------------------------
Slides: http://www.ogf.org/OGF33/materials/2350/NML+Progress+OGF33.pptx
(slides 14-18)
URL:
http://forge.gridforum.org/sf/tracker/do/listTrackers/projects.nml-wg/tracker

Freek added tasks in the Tracker at the Gridforge website.

Everyone is invited to shepherd any item that is marked as "New" or
"Waiting for Volunteer".

Jeroen notes that OWL is now the more standard syntax for ontologies.
Proposes to change RDF to OWL.

Jerry is volunteering to review the schema document once it nears
completion.

Jeroen volunteers to write the OWL schema
Jason volunteers to finish the RNC schema
Freek volunteers to help with the schemas

Jason Zurawski - XML syntax
---------------------------
Jason energetically led the working group through a couple of examples
in XML, trying to reach consensus on syntax. The effort started in the
hallway, and continued during the third session.

The XML files are available from subversion repository at
http://forge.ogf.org/svn/repos/nml-examples/20110922

Consensus was reached on the single layer XML syntax.

Freek noted that he still thought the syntax of NML relations could be
improved, but also made it clear that there is rough consensus on
<nml:relation type="name_of_the_relation"/>. He reserved the right to
re-iterate on a previous proposal after creating an implementation, to
show the advantages (if any), but also urged the working group to
continue with the current syntax, and not wait on that proposal.

The goal of the third working session was to create a syntax to describe
a VLAN switch, including a 802.1x-capable link between two devices.

This problem was tackled in the following steps:

1. nmlbase.xml: A syntax to describe channels over a single link, in
this case a single (default) VLAN (ID 0) over a link between Houston and
Atlanta. This introduces two complimentary concepts:
  - A "providesChannel" relation between the (physical) link and
(channel) link.
  - A syntax to describe adaptation between a (physical) fibre port, and
the (logical) VLAN ports. This was done using an "adaptation" Service,
which includes the adaptation between the two ports.

2. switch_matrix_without_labels.xml: Introduce the concept of a
"switching" Service, that can provide cross connects.

3. switch_matrix_w_labels.xml: Introduce the concept of "labels"
associated with ports (the possibility if they can be associated with
links as well was not discussed. This is still open for debate.)

4. configured_vlan.xml: Description of a single VLAN as a "switching"
Service.

5a. 2vlans_option1.xml: Description of two VLANs as two "switching"
Services.
5b. 2vlans_option2.xml: Description of two VLANs as one "switching" Service.
  - It should be decided which of these two option is the best syntax.
  - 2vlans_option1.xml is more verbose, but easier to parse
  - 2vlans_option1.xml scales better (more compact syntax), but harder
to parse

The XML files are a first pass on agreeing with the syntax. They are not
yet production ready examples:
- they contain comments listing some (usually minor) open issues
- some agreed upon syntax is commented out, so it validates against the
RNC schema
- some syntax may be superfluous for a given example, the examples
should be trimmed to remove redundant code.

Freek, Jason and Jeroen agreed to work on these examples, so they can be
used in the schema document (or in an accompanying document).