15:00:41 <dpertin> #startmeeting fog-edge-massively-distributed-clouds
15:00:45 <openstack> Meeting started Wed Apr 25 15:00:41 2018 UTC and is due to finish in 60 minutes.  The chair is dpertin. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:46 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:48 <openstack> The meeting name has been set to 'fog_edge_massively_distributed_clouds'
15:00:52 <dpertin> #chair dpertin
15:00:53 <openstack> Current chairs: dpertin
15:00:58 <dpertin> Hi here!
15:01:08 <dpertin> #topic roll call
15:01:33 <dpertin> #info dpertin - dimitri.pertin@inria.fr
15:01:48 <dpertin> Waiting few minutes for people to come
15:01:49 <rcherrueau> #info rcherrueau - Ronan-Alexandre.Cherrueau@inria.fr
15:01:56 <ansmith> o/
15:02:05 <marie_> Hi
15:02:14 <dpertin> During this time you can jave a look at the agenda in the pad and contribute by adding any content you would like to talk during this meeting
15:02:22 <dpertin> #link
15:02:43 <dpertin> #link https://etherpad.openstack.org/p/massively_distributed_ircmeetings_2018 (line 466)
15:03:59 <dpertin> Our beloved chairman _ad_rien is not here today since he enjoys a well-deserved rest due to the holiday period in France
15:04:42 <dpertin> I volunteered to chair this meeting but I would appreciate any help since it is my first time ;)
15:06:09 <dpertin> Ok, let's start!
15:06:18 <dpertin> #topic announcement
15:06:49 <dpertin> From our side, we released a new version of EnOS, our tool to deploy and benchmark different configurations of OpenStack
15:07:04 <dpertin> #link https://github.com/BeyondTheClouds/enos/releases/tag/3.0.2
15:07:12 <dpertin> ^ here's the release note
15:07:31 <prmachan> #info prachan
15:07:37 <dpertin> While this version is based on Pike, a new version should be released soon for OpenStack Queen
15:07:39 <prmachan> #info pramchan
15:08:02 <dpertin> #link https://github.com/BeyondTheClouds/enos/pull/239
15:08:32 <dpertin> I think that's all from our side regarding announcements, does anyone want to add something?
15:09:13 <msimonin> o/
15:09:16 <prmachan> #yes
15:09:40 <dpertin> prmachan: please go ahead ;)
15:10:46 <prmachan> #you are talking of enos and is there any one who can present that (Adrien?) at edge cloud meeting Thursday 12.30-1.45 PM at Vancouver summit?
15:11:29 <dpertin> prmachan: do you have a link please regarding this meeting?
15:11:56 <prmachan> #After last presentaion at was it Sydney I do not recolect new upates presented and we are looking form Edge perspective to use enos
15:13:03 <prmachan> I have added to PTG let me get etherpad and have asked Alan/Castri to add it if Adrien can represent at Summit or anyone he assigns to
15:14:08 <dpertin> prmachan: this is not scheduled now but you can add a new item here:
15:14:11 <dpertin> #link https://etherpad.openstack.org/p/FEMDC_Vancouver_brainstorming
15:14:39 <dpertin> Are you refering to the FEMDC F2F on Thursday?
15:14:56 <prmachan> https://www.openstack.org/summit/vancouver-2018/summit-schedule/events/21685/edge-working-group-session
15:14:56 <dpertin> #link https://www.openstack.org/summit/vancouver-2018/summit-schedule/global-search?t=femdc
15:15:52 <prmachan> I have requested an additional session to cover following the onle listed above and room numbet will be smae except seesion 2 of edge will cover cross project co-ordination
15:16:35 <prmachan> and I believe Adrian and MPP team is very improtant to to present enos to test different edge cloud uses cases
15:17:21 <dpertin> prmachan: can you add an item regarding enos presentation/update in the brainstorming pad? Since we have two sessions we might have time for that
15:17:51 <prmachan> #plus ONAP & OPNFVedge group are also requested to join the discussions to get blueprints discussions on minimizing overlaps
15:18:04 <msimonin> prmachan:  are you expecting anything specific about such presentation  ?
15:18:15 <prmachan> Yes I have already added and request Adrien to help us by providing support
15:19:16 <rcherrueau> prmachan: Are you using EnOS, or do you plan to use it?
15:19:21 <prmachan> #more an interactive session with some presentaion and co-ordinating with Dublin PTG discussiosn to get internal as well upstream co-ordination
15:20:03 <prmachan> #Want to see if we can use EnoS for testing edge clusters as part ofEdge cloud
15:21:14 <prmachan> #info Session 1 : Can focus on OpenStack internal efforts on Edge Cloud Co-ordination
15:21:55 <msimonin> This would be maybe more a feataure requests list to EnOs to better support testing of edge clusters
15:22:05 <prmachan> #info Thursday May 24 Session2:12.40-1.50pm, in Room 217  ( Cross-Project Blueprints/Specs ONAP/OPNFV/OpenStack/TIP to minimize overlaps)
15:23:02 <prmachan> #Note the context is Openstcak Vancouver summit if you missed that
15:24:00 <prmachan> #msimoni has captured and summarized it very well - Thanks
15:24:13 <msimonin> #thanks :)
15:25:11 <dpertin> Noted, thanks, the list of items regarding the Vancouver summit is still open, let's wait few weeks before setting the final agenda of the sessions
15:25:32 <dpertin> Ok, shall we move forward? Other annoucements?
15:26:03 <dpertin> Let's move to the next topic
15:26:07 <prmachan> #sure dpertin our goals are clear and we can bring more precision to it with all your help
15:26:26 <dpertin> #topic ongoing-topic-gapanalysis
15:26:43 <dpertin> From our side, we are still waiting for the HotEdge notification
15:27:12 <dpertin> Good news? Bad news? Let's keep our fingers crossed ¯\_(ツ)_/¯
15:27:30 <dpertin> Now, regarding the gap analysis...
15:28:05 <dpertin> Thanks to csatari who organized two meetings, we have started to review the wiki page he created following the discussions at the PTG
15:28:27 <dpertin> #link https://wiki.openstack.org/wiki/OpenStack_Edge_Discussions_Dublin_PTG
15:29:03 <dpertin> csatari: don't know if you are here, would you like to add anything about it?
15:29:28 <dpertin> A third meeting will be scheduled soon if you want to be part of the review process
15:30:07 <prmachan> +1
15:30:42 <dpertin> Following this analysis, we are some people at Inria discussing about how to extend OpenStack to provide the "collaboration between edge cloud instances" feature as defined in the wiki page made by csatari
15:31:12 <dpertin> "cloud instance" refers to independent openstack instance like the ones provided by the OpenStack region mechanism
15:31:43 <dpertin> For that, we started to list few simple operations like: booting a VM on Site A, using an image from Site B
15:31:48 <prmachan> reviewed it yesterday and  which profile Smalla or Medium you use here in MPP/enos?
15:32:23 <rcherrueau> prmachan: MPP?
15:32:59 <dpertin> From this list, we identified that cross-site operations can be either inter-service/project (e.g. my previous example involves nova from site A and glance from site B) or intra-service/project (e.g. list VMs on site A and site B uses nova from site A and nova from site B)
15:33:27 <prmachan> So what they call Minimum Viable Solution(MVS) , how enos will use it and will conpute node or sites focus on multi core Massive Parallel Processing?
15:33:47 <prmachan> I see
15:33:50 <dpertin> This has an impact regarding the communications (apis, bus, databases) involved during a collaborative operation, which is critical regarding our analysis
15:35:43 <prmachan> Sounds perfect, question is what do I make of Original site vs. remote site therin , des it refer to Central and Edge sites/nodes?
15:36:35 <prmachan> so when you say Inter and Intra does refer to edge to edge or edge to central?
15:36:36 <dpertin> For instance: for the inter-service example, nova from A asks an image to the glance from B, and glance can find the information in the database of B - however for the intra-service operations, nova in B might look for information only available in the database of A
15:37:18 <dpertin> prmachan: I would more segment operations between mono-site operations (either local or remote) and collaborative operations
15:37:59 <prmachan> Are we tallking peer to peer openstack operations here independent of hierarachy of central and edge clouds?
15:38:55 <prmachan> Now I undertans , thanks for expalinig mono-site
15:38:57 <dpertin> collaborative operations = operations that involve multiple sites - it can be based on P2P mechanism
15:39:25 <rcherrueau> prmachan: regarding EnOS, there is nothing settled. EnOS helps us to deploy several OpenStack configuration for performance analyses. At some point, EnOS should be able to deploy small or medium edge nodes (to be honest, I am not sure to understand what small and medium refer to, in this context)
15:39:51 <dpertin> Our questions now is should we implement things at the API level, at the low-level (DB,bus) or in-between
15:40:23 <prmachan> Those are more hardware profiles and can be atbest described through topology templates for hardware
15:40:42 <dpertin> Furthermore, such question were probably met by the people who designed nova cells and/or projects like tricircle/trio2o, so having a look at those projects is probably a good way to fuel our analysis
15:41:20 <dpertin> Maybe we should go forward and come back to it later, or on the pad
15:41:24 <dpertin> xx:41
15:41:38 <prmachan> ok
15:41:50 <dpertin> #topic ongoing-topic-cockroach
15:41:59 <dpertin> marie_: rcherrueau: the floor is yours ;)
15:42:01 <rcherrueau> prmachan: Yes you are right, you can imagine a fancy DSL to describe your topology and it could be could to get such DSL in EnOS. But, this is not in the roadmap
15:42:41 <marie_> We redefined a better test plan last week
15:43:17 <marie_> The one written in the pad
15:44:09 <marie_> The main difference is we are now testing the influence of regions separated by a high latency, but the regions itself have several nodes that have no latency between them
15:44:21 <dpertin> marie_: nice, is your test plan described in the performance team documentation?
15:44:22 <marie_> *we are also testing
15:44:49 <marie_> not yet, I have to rewrite that also, since the test plan has changed
15:45:13 <msimonin> What is "Number of nodes" ?
15:45:28 <rcherrueau> Number of nodes in the database cluster
15:46:01 <msimonin> what is the quorum size ?
15:46:10 <dpertin> marie_: any results from this test plan?
15:46:19 <marie_> it's the default size
15:46:22 <rcherrueau> msimonin: 3 for cockroach (default)
15:46:39 <prmachan> #is there a link to follow the test plan?
15:46:44 <marie_> we have -mostly- the results for the two first goals
15:46:53 <msimonin> rcherrueau: thanks
15:47:02 <dpertin> marie_: ok thanks (regarding the perf wg)
15:47:07 <marie_> and we are actively working on the third part of the experiments
15:48:11 <msimonin> 1*3 means 1 site with 3 db nodes ?
15:48:22 <marie_> yep
15:48:30 <marie_> and no latency between them
15:49:26 <dpertin> marie_: have you tried the last new version of cockroach?
15:49:41 <dpertin> I heard they made improvements regarding the performance
15:49:53 <dpertin> s/last//
15:50:08 <marie_> Yep, we patched the framework to use Cockroach 2
15:50:51 <dpertin> anything else from your side?
15:51:01 <marie_> I don't think so
15:51:21 <rcherrueau> yes, I add something lines 504-512
15:51:36 <rcherrueau> Some advocate keystone is write once, multiple multiple multiple ... reads at database level
15:51:36 <rcherrueau> 
15:51:39 <marie_> oh, right, thanks !
15:52:08 <rcherrueau> And it could be cool to know the real ratio because this may have an impact when you choose a replication strategy or another
15:52:31 <rcherrueau> so here are some ratio we compute with the tool marie_ developed.
15:52:39 <msimonin> great ! :)
15:52:49 <dpertin> Is it computed from rally tests?
15:53:20 <rcherrueau> Note that the ratio of reads/writes may be a little more in favor of reads, than what it is presented here because, the following also take into account the creation/deletion of rally context that performs INSERT and DELETE.
15:53:33 <rcherrueau> dpertin: yes from rally scenarios
15:53:46 <dpertin> Ok let's continue this topic on the pad, please
15:53:51 <dpertin> #topic ongoing-topic-amqp
15:54:32 <dpertin> msimonin: ansmith: any update from your side?
15:55:04 <ansmith> I defer to msimonin as kgiusti could not attend today
15:55:22 <msimonin> Hum nothing specific, we are starting to look at locality deliveries when distributing the agents
15:57:24 <rcherrueau> cool
15:57:42 <dpertin> Ok 2 minutes left
15:57:51 <dpertin> #topic vancouver-summit
15:58:00 <dpertin> We have an etherpad regarding the topics that should be discussed in Vancouver
15:58:07 <dpertin> #link https://etherpad.openstack.org/p/FEMDC_Vancouver_brainstorming
15:58:31 <dpertin> I think that from our side, we will have more elements regarding the missing parts in the key components in the next few weeks
15:58:50 <dpertin> Does anyone want to add something?
15:59:35 <dpertin> Thanks guys! I was glad to chair this session :)
15:59:45 <dpertin> If you have any comment, please feel free to add them into the pad
15:59:51 <dpertin> #endmeeting