Info

Information Services (IS) Working Group
February 02, 2009, 4:10 PM - 5:10 PM
CST (UTC-6)
Location: MSC: 229/230

Session Info

To Dial In:
Call: +1-734-615-7474 (PSTN CALL-OUT DOES NOT WORK)
or +1-866-411-0013 (toll free US/Canada Only)
Enter access code: 0145258

Agenda

  • Choose time for a periodic conference call.
    • the poll
    • How often to meet (Bi-weekly, Monthly)
  • Current Activities
    • Discuss some recent developments in the IS effort
    • Open forum on group direction
  • Develop/Discuss Charter
  • Discuss Requirements / Architecture document

Attendees

  • Martin Swany (Univ. Delaware)
  • Jason Zurawski (Internet2)
  • Aaron Brown (Internet2)
  • Jeff Boote (Internet2)
  • Andy Lake (Internet2)
  • Phil Demar (Fermilab)
  • Carla Hunt (MCNC)
  • Brian Tierney (ESnet)
  • Walt Prue (USC)
  • Rich Santillo (Penn State)
  • Michael Contino (Penn State)
  • Joe Metzger (ESnet)

Notes

Working Group Preliminaries

Opening discussion focused on the purpose and scope of the working group. Some questions from the group regarding how this group fits in with:

  • Other NTAC working groups
    • DCN - Recommendations here will dictate how DCN components register existence and capabilities with IS software. Also how client software (Phoebus/Lambda Station/Terrapaths) can take advantage of this registration for service location.
    • Performance - IS directly affects perfSONAR services, particularly the performance nodes.
  • OGF Working Groups
    • NML - Topology representation defined here will drive how the IS shares and distributes this info.
    • NSI - Topology information will be defined here as well.
    • NMWG - No direct relationship
    • NMC - Communication structures and exchange rules will be defined here.

Additional questions were raised regarding deliverables and WG activities. In the end a consensus was reached where the group should aim to produce documents (White papers, technical reports) that make recommendations for other NTAC working groups. Technical input (e.g. schema definition, etc.) will come from the OGF working groups as many of the members of this WG are also members of the OGF working groups.

IS-WG members should vote in the doodle poll to decide on an appropriate time for a WG meeting. It is proposed that the group meet Bi-weekly but this can be discussed on the first call.

Summary of perfSONAR IS Activities

Jason gave an update on the status of the perfSONAR development and deployment status of the service formerly known as the perfSONAR Lookup Service. Slides are available here .

Discussion points here consisted of:

  • Is there any guiding force used in the selection/use of the keywords discovery paramter.
    • Currently no, sites are free to use whatever they choose
    • This can lead to a explosion in the keywords space
    • Suggestions for addressing this:
      • Configuration tools should have a way to display or suggest keywords
      • Use formatting tricks to eliminate off by one errors (e.g. capitalization or spaces)
  • In addition to APIs, provide command line, web, and documentation to using the IS. This will increase adoption
  • Proceed carefully with making registration API available to outside services.
    • This topic will be discussed in the next section.
  • Combination of the legacy Topology Service (currently designed similar to CNIS requirements) and the (Global|Home) Lookup Service to provide a unified interface to network information

Potential Directions (TS/LS combination)

It is on the roadmap for perfSONAR developers to combine the insides of the legacy Topology Service with the Lookup Service to provide distributed topology information. There are open questions that effect this transition, as well as the potential interactions that perfSONAR services, measurement tools, and outside services such as DCN and phoebus.

Martin discussed a potentially troublesome issue regarding the registration of node elements - each service needs to be aware of this concept to prevent duplicate registrations of the same information particularly spanning services. Each service must have the ability to locate the proper node from the IS (if it is present) or register a proper node definition for other services to share. Currently the registration API does not account for this issue, so before future releases this must be addressed.

A side discussion was started on how to properly identify a service without the aide of identifiers (such as NURNs, discussed below). The concept of using host certificates (self signed perhaps) as the way to identify a host/node when registering information about a service and how it connects to topology. This will be explored along with UUIDs, unix host identifications standards, and similar technologies.

Potential use for the combination service will be:

  • Closest Resource Problem
    • Find the closest perfSONAR service
    • Find the closest measurement daemon along a path
  • Pathfinding
    • DCN specifically, other services in general
  • Replica Selection
  • Topology Oracle

Potential Directions (NURNs)

An issue related to the previous discussion topic is the use of unique and useful identifiers. This is still an active discussion topic in larger groups such as NML, NSI, and GLIF. There are two proposals:

  • Structured NURNs
  • Unstructured IDs
    • Blobs of data that do not contain any predictable way of identifying what the content they describe may be

A compromise was discussed, instead of:

  • urn:ogf:network:domain=internet2.edu:node=packrat:port=eth0

Use something similar to:

  • urn:ogf:network:domain=internet2.edu:id=438247239847

This will be discussed in the appropriate forums in the future.

Next Meeting

The next meeting will be held after the result of the doodle poll.

The next face to face meeting is tentatively scheduled for the Internet Spring Member Meeting

  • No labels