Date
1 - 17 of 17
Agenda for 4/29, 1p PT
|
Anees A Shaikh <aashaikh@...>
For this Monday's call, agenda and call-in information are posted on the
wiki here: https://wiki.opendaylight.org/view/Tech_Work_Stream:Main This week: Discussion and demo of model-driven approaches for an extensible southbound protocol abstraction layer. Jan Medved, Tony Tkacik (Cisco) thanks. -- Anees ------- Monday, April 29, 2013 1:00 pm, Pacific Daylight Time (San Francisco, GMT-07:00) Event number: 665 676 801 Event password: odp1 Event address for attendees: https://linuxfoundation.webex.com/linuxfoundation/onstage/g.php?d=665676801&t=a ------------------------------------------------------- Teleconference information ------------------------------------------------------- US TOLL: +1-415-655-0001 Access code: 665 676 801 |
|
|
|
Yi Yang
Will the meeting be recorded?
toggle quoted message
Show quoted text
Yi On 4/28/13 3:25 PM, Anees A Shaikh wrote:
For this Monday's call, agenda and call-in information are posted on the |
|
|
|
Anees A Shaikh <aashaikh@...>
Yi, I'm not sure about recording (will have to see if it's possible on
Webex config) , but we will post all the presentation material. thanks. -- Anees discuss-bounces@... wrote on 04/28/2013 09:29:38 PM: From: Yi Yang <yyodl1999@...>the wiki here:d=665676801&t=a |
|
|
|
Phil Robb
Hi Yi and Anees. Recording should be possible, but we have not yet done it on this Webex config. Anees, please remember to announce that we are recording the call at the beginning, and I'll get it turned on. Thanks, Phil. On Mon, Apr 29, 2013 at 10:45 AM, Anees A Shaikh <aashaikh@...> wrote: Yi, I'm not sure about recording (will have to see if it's possible on Phil Robb Director - Networking Solutions The Linux Foundation (O) 970-229-5949 (M) 970-420-4292
Skype: Phil.Robb |
|
|
|
Yi Yang
Thanks, Anees and Phil!
Yi On 4/29/13 1:18 PM, Phil Robb wrote:
|
|
|
|
Chris Wright <chrisw@...>
* Anees A Shaikh (aashaikh@...) wrote:
Discussion and demo of model-driven approaches for an extensiblePlease send the slides out |
|
|
|
Rob Sherwood
Actually, if you could send the slides out _now_ it would be appreciated. I want to rewind a bit on the call and try to ask a few questions online. TIA,
- Rob . On Mon, Apr 29, 2013 at 1:17 PM, Chris Wright <chrisw@...> wrote: * Anees A Shaikh (aashaikh@...) wrote: |
|
|
|
Colin Dixon <ckd@...>
I just wanted to clarify one point. |
|
|
|
Rob Sherwood
Thanks to Jan and all for the presentation -- it certainly helps clarify what a model driven SAL means. That said, it looks _a lot_ like the BigDB work[1] that I demoed a while back. All of the APIs are generated from YANG, different modules are providing different parts of the subtree, there is no base-line translations between south bound APIs (but they could be layered on top by other drivers).
Am I misunderstanding? Hopefully someone can help clarify this for me. Thanks in advance, - Rob .
[1] @Anees: I went looking for the links to the BigDB talk that I gave -- can you share where they ended up? I didn't see them on https://wiki.opendaylight.org/view/Tech_Work_Stream:Main which is where I expected to find them. Thanks in advance!
On Mon, Apr 29, 2013 at 2:10 PM, Colin Dixon <ckd@...> wrote:
|
|
|
|
Rob Sherwood
FYI, Anees mailed me the BigDB links: (Thanks Anees!) - Rob . On Mon, Apr 29, 2013 at 3:38 PM, Rob Sherwood <rob.sherwood@...> wrote:
|
|
|
|
Chris Wright <chrisw@...>
* Rob Sherwood (rob.sherwood@...) wrote:
FYI, Anees mailed me the BigDB links:Cool, perhaps you could grab it and add it to the current wiki for those that don't have credentials to the collabdocs. thanks, -chris |
|
|
|
David Erickson <derickso@...>
Is the recording available for today's
meeting? I saw a link to the slides is available
(https://wiki.opendaylight.org/view/Tech_Work_Stream:Main) but no
Webex recording yet.
Thanks, David On 4/29/2013 10:18 AM, Phil Robb wrote:
|
|
|
|
Phil Robb
Hi David: The link to today's OpenDaylight Technical Work Stream recording is now posted on the wiki. For convenience, I've also provided it here: https://linuxfoundation.webex.com/linuxfoundation/lsr.php?AT=pb&SP=EC&rID=6781737&rKey=259ae7e837db1b87
Phil. On Mon, Apr 29, 2013 at 7:42 PM, David Erickson <derickso@...> wrote:
Phil Robb Director - Networking Solutions The Linux Foundation (O) 970-229-5949 (M) 970-420-4292
Skype: Phil.Robb |
|
|
|
Anees A Shaikh <aashaikh@...>
Thanks Phil !
Folks, please note that slides and recording from today's call are now available on the wiki, as well as pointers to a related talk on BigDB that Rob Sherwood gave a few weeks back. https://wiki.opendaylight.org/view/Tech_Work_Stream:Main -- Anees discuss-bounces@... wrote on 04/29/2013 10:50:53 PM: From: Phil Robb <probb@...> On Mon, Apr 29, 2013 at 7:42 PM, David Erickson <derickso@...>wrote: Is the recording available for today's meeting? I saw a link to thehttps://wiki.opendaylight.org/view/Tech_Work_Stream:Main ) but no Webex recording yet.on.
On Mon, Apr 29, 2013 at 10:45 AM, Anees A Shaikh <aashaikh@...>wrote: Yi, I'm not sure about recording (will have to see if it's possible onFrancisco, _______________________________________________GMT-07:00)d=665676801&t=a
_______________________________________________
|
|
|
|
David Erickson <daviderickson@...>
Thanks Phil!
On 4/29/2013 7:50 PM, Phil Robb wrote:
|
|
|
|
David Erickson <daviderickson@...>
I just finished watching the demo (had
to leave the call early), this may be a small point, but one thing
I noticed was the API version was being embedded in the package
name, I am curious how this is intended to work with non-generated
code that depends on the package names remaining the same?
-D On 4/29/2013 8:27 PM, David Erickson wrote:
|
|
|
|
Jan Medved
Rob,
sorry for the late response, I was not subscribed to the discuss mailing list, and thus did not see your and others' requests/emails. Please see inline, and also find attached the presentation from yesterday.
Please see below.
Your presentation stated that BigDB generates both REST and Java APIs. I am not sure what exactly gets generated for REST, but from your demo I think the way clients will access REST APIs will be similar in both the SAL and BigDb. We would love
to see what you do for Java API generation - as you saw in the demo we made a lot of progress, and would definitely like to compare notes with you.
Now, while in the model-driven SAL Java APIs are generated at compile time, the Controller (the SAL) generates code for Java API Providers at runtime. The generated code translates between the binding-independent format and the Java binding-aware
format (iow, code is generated for Java bindings).
I haven't heard you mentioning runtime code generation concept in the BigDB talk. How do you guys plan to address the code implementing the APIs? It's an interesting topic that we should discuss.
Note also, that if a provider (the entity that sources the model) and a consumer can both use the binding-independent data format (for example, if both the provider and the consumer are netconf plugins), there is no need for any API/code generation.
Yes, in both BigDB and in the model-driven SAL different yang modules can provide different parts of the network tree. I don't know where the yang modules come from in BigDB, and how they're loaded into the runtime (the demo was inconclusive).
In the SAL, yang modules can be provided by controller plugins or by network elements (if supported by the underlying SB plugin, such as Netconf). With OSGI, plugin modules can be dynamically loaded/unloaded at runtime. Both these requirements
mean that yang models need to be dynamically loaded into the SAL at runtime and the SAL must be able to dynamically "stitch" together the overall network model tree at runtime.
I don't know how you are putting together the network tree (compile time? runtime? do you need to replace the whole tree at once, or do you dynamically add/delete subtrees?), but we found that in order to create and maintain the tree dynamically,
we had to introduce the 'mount' concept into yang - see https://datatracker.ietf.org/doc/draft-clemm-netmod-mount/.
In addition to being able to dynamically modify the network model, mounting allows for the network elements' models to be used as-is in the controller. That means, if a new feature shows up in a network element or a new network element shows
up in the network, the controller can just load it's model, mount it in the appropriate location in the network tree and then expose it to NB apps that understand the network element's new model (Note that for binding-independent consumers no code/API generation
is needed, for binding-aware consumers the API is generated at compile time, then loaded as an OSGI bundle into the controller, which will auto-generate all the required code).
Also, there is no need to create a 'shadow' model of the network element model for the use in the controller.
Anyway, if you don't use the yang 'mount', there are likely some differences between BigDb and the SAL :-) Again, would love to compare notes.
Anyway, that's by design. There are many use cases where an app needs untranslated access to network elements. As you pointed out below, if an abstraction is needed, we can define a model for the abstraction, then create a plugin (an OSGI bundle)
that will translate between the 'raw' network element model and the abstraction. The point is that right now we don't know what kind of abstractions will be needed (I would guess they would be different in different environments anyway), so we need to create
a framework where we can define, implement and deploy them relatively easily.
There are couple other architecture/design aspects that may be different (I am guessing based on your talk, not knowing the internals of BigDB):
Tony will certainly provide more when he comes online :-)
Hope the above helps.
/Jan
|
|
|