#opendaylight-meeting: tws

Meeting started by colindixon at 17:02:30 UTC (full logs).

Meeting summary

  1. agenda (colindixon, 17:02:35)
    1. today Michal Polkoráb will present info on the topology processing framework (colindixon, 17:03:37)
    2. if you have other topics, please let Keith and I know, and we’ll get you on the agenda (colindixon, 17:04:40)
    3. https://wiki.opendaylight.org/view/Tech_Work_Stream:Main#Meeting_Schedule_and_Logistics (colindixon, 17:04:50)

  2. topology processing framework (colindixon, 17:04:58)
    1. restarted recording around 10:07, and we’re starting things in case you’re trying to sync meeting minutes and the recording (colindixon, 17:06:35)
    2. the topology procesing framework sits in the controller (colindixon, 17:07:02)
    3. its primary role is to create aggregated topology interfaces instead of one (or more) topology per protocol (colindixon, 17:07:40)
    4. right now it works by creating logical nodes and embedding them in a single logical topology (colindixon, 17:08:03)
    5. in addition to logically merging topologies, topologies can be filtered so that you can see a subset of topology (colindixon, 17:10:12)
    6. right now there is a way for you to provide a Comparator (I think) which will say if two nodes should be unified when aggregated topologies are created (colindixon, 17:12:20)
    7. there is also a way to filter based on the Node IP (colindixon, 17:12:28)

  3. live-view operations (colindixon, 17:12:53)
    1. the framework appropriately handles and produces node added, node removed, and node changed (colindixon, 17:13:56)
    2. https://wiki.opendaylight.org/view/Topology_Processing_Framework:Developer_Guide the developer guide (colindixon, 17:14:01)
    3. https://wiki.opendaylight.org/view/Topology_Processing_Framework:Main the project main page (colindixon, 17:14:15)

  4. equality aggregation use case (colindixon, 17:15:06)
    1. they show the 80-topoprocessin-confg.xml file (colindixon, 17:16:28)
    2. shows the running ODL karaf container: the bgp features are installed, the topology processing feature is installed, and the restconf feature is installed (colindixon, 17:17:57)
    3. they push a REST request telling topoprocessing to merge the 3 bpg topologies into one (colindixon, 17:18:43)
    4. show via a REST get that there is now a logical topology (colindixon, 17:19:06)
    5. they push 3 nodes with the same IP address into the 3 different pcep (bgp) topologies and show that they get aggreged into the same node in the logical topology (colindixon, 17:20:04)
    6. they change the IP address of one of the nodes so that it has a different IP addess, this is reflected in the logical topologies (colindixon, 17:21:17)
    7. they also demonstrate remving nodes (colindixon, 17:21:26)
    8. the key point here is that in the equality use case only published nodes that can be aggregated into the logical topology, if there are nodes which have no potential aggregations, they aren’t listed (colindixon, 17:22:36)

  5. unification aggregagion use case (colindixon, 17:22:53)
    1. in this case, the logical topology will contain all nodes in the topologies it is reading from even if there is no way to aggregate it with anything else (colindixon, 17:23:27)
    2. colindixon asks if there’s any reason you’d use the equality use case over the unification, the answer is that seems likely but for now they suppor both because it wasn’t hard (colindixon, 17:24:33)
    3. they go through the demo of adding, changing, and removing nodes and showing they are appropriately represented (colindixon, 17:25:00)
    4. colindixon asks how it finds the fields to compare equality, tony asks Michal to show the network topology configuration to answer the question (colindixon, 17:27:20)
    5. the config shows that they are specifying the fields to compare for equaity, but they have to be fields defined as augmentations to topoglogy node (colindixon, 17:29:39)
    6. critically, that means you can’t corelate OF nodes since they augment inventory nodes (colindixon, 17:31:17)
    7. colindixon asks abhijitkumbhare if there’s any plan to move OF from inventory to topology, he says no, not at this time (colindixon, 17:31:38)
    8. abhijitkumbhare notes that we can discuss this during the Beryllium release plans (abhijitkumbhare, 17:33:02)
    9. michal says he hopes to work on fixing that during Beryllium (colindixon, 17:33:48)
    10. colindixon asks if we could use use comparators instead of just equality, micahal says that’s planned for Beryllium along with some scripting (colindixon, 17:35:35)

  6. filtering use case (colindixon, 17:36:06)
    1. Micahal says that they have IP address filtering only right now, but he’s happy to implement whatever else you want if you ask (colindixon, 17:36:33)
    2. the demo shows that the logical topolgoy, does in fact, filter based on IP addresses (colindixon, 17:37:48)
    3. colindixon asks if you could extend this to be able to add new filters after compile time (colindixon, 17:39:23)
    4. Michal says not now, but maybe in Be (colindixon, 17:39:30)

  7. wrapping up (colindixon, 17:39:34)
    1. currnetly this works with real topology and real data, as long as everything augments the topology model (colindixon, 17:39:57)
    2. longer-term goals multi-layer multi-topology support (colindixon, 17:40:14)
    3. scripting support for programmable rules in filtering, equality and unification (colindixon, 17:40:39)
    4. LuisGomez asks, you showed nodes, what about links and termination points (colindixon, 17:41:10)
    5. Michal says links and TPs are aren’t supported, but they’re hoping to do that in Be (colindixon, 17:41:26)
    6. colindixon asks what they used to make the slides, the answer is impress.js (the slides were cool) (colindixon, 17:43:03)
    7. they will post the slides to the wiki (colindixon, 17:43:15)
    8. colindixon says that this is really cool and is important for anyone writing an app that wants be support multiple different southbound protoocols and deal with toptology in a protocol-independent way, because you will have to build your own version of this or use this stuff (colindixon, 17:49:22)
    9. hideyuki asks what colindixon meant above about the inventory model vs. topology model (colindixon, 17:49:45)
    10. colindixon says that for historical reasons, we have two top-level models with nodes: inventory (which is ODL-specific) and topology (which is from an IETF RFC) (colindixon, 17:51:06)
    11. for historical reasons the OF plugin uses the inventory model, while pretty much everyone else uses topology and we should really be moved to the topology model (colindixon, 17:54:37)

  8. asciidoc stuff (colindixon, 18:01:46)
    1. tony shows how to do some automated translation of wiki to asciidoc (colindixon, 18:03:51)
    2. tony will send an e-mail to the mailing list describing what he’s done (colindixon, 18:04:59)
    3. there was a lot of discussion around how to improve docs going forward (colindixon, 18:05:11)
    4. dbainbri asks if we want to have a model which is AsciiDoc as the primary and that’s where people edit it or if we want to move to a mode where the wiki is the primary and we simply use AsciiDoc to archive things at a release (colindixon, 18:05:46)
    5. several people say that having HTML would be much better than PDFs (colindixon, 18:05:59)
    6. colindixon says that we have HTML, it’s just way too segmented (one document per subsection or subsubsection) to be useful (colindixon, 18:06:52)
    7. tony shows that on github, we can use some of their asciidoc renderer to get HTML views of the live docs in the repo (colindixon, 18:07:16)
    8. https://github.com/opendaylight/docs/blob/master/manuals/getting-started-guide/src/main/asciidoc/general_installation.adoc this is a good example (colindixon, 18:08:41)
    9. odlcasey asks about how the documentation effort is staffed, colindixon says it’s staffed by people who mostly have day jobs as well and thus we just don’t have the raw documentation horsepower except from the projects themselves (colindixon, 18:10:08)
    10. colindixon invites anyone who wants to help to come to the docs project meetings and plan Be features if they are interested (colindixon, 18:10:35)
    11. https://wiki.opendaylight.org/view/Documentation this page has information on documentation meetings and plans (colindixon, 18:11:09)

Meeting ended at 18:11:14 UTC (full logs).

Action items

  1. (none)

People present (lines said)

  1. colindixon (75)
  2. odl_meetbot (7)
  3. abhijitkumbhare (3)
  4. hideyuki (2)
  5. dbainbri (2)
  6. gzhao (1)

Generated by MeetBot 0.1.4.