Page History

...

Value measurement formats

Monday,  April 6 - Client Implementation, Vehicle Data

First call - Asia friendly time 11:30am CET

Agenda

• weekly sync with Sanjeev on the client implementation

Participants

• Sanjeev, Gunnar, Philippe

Minutes

proof-of-concept design

• Gunnar; presents the refinement made on the component interface last week, in CCS Proof-Of-Concept - Work Breakdown Structure
• Gunnar: we are now collecting proposals for graphql schemas
• discussion on impact on Sanjeev's workitems
• Sanjeev: in my opinion, we might need someting like an object store for the data lake
• Gunnar: Ulf is likely to propose a relational data base
• Philippe: Ulf will make a presentation on the data lake in this afternoon CCS call
• Sanjeev: is there any reason to use a relational database ? I saw the data coming as a continuous data stream, unless there is a need of classification of data, I am wondering why we should use a database
• Gunnar:  understood, in the long run, the name data lake means there could be very diverse data and it could make more sense to look into an object oriented database for instance
• Sanjeev: I stumbled in similar problems in the IoT domain
• Gunnar: we might go to Hadoop in a later version of the proof-of-concept

client implementation

• Sanjeev: last week worked with Ulf on fixing the Gen 2 server, we have a permanent node for the server provided by MIT, this took all the time I could spend on the project
• Sanjeev: do I have permission to create a repo in  GENIVI github ?
• Gunnar:  do we create a separate repo for the client ?
• Sanjeev: in my opinion, the W3C server module will be become a submodule of the CCS client I will be working on
• Gunnar: goes to github/genivi/ccs-w3c-client repo created, Sanjeev set up as admin of the repo

Second call - US friendly time 16:00 CET

Agenda

• Data Lake proposal
• Data Categories, VSS tree update with EV signals (skipped since Benjamin is not attending today)
• Value measurement formats
• Sprint & backlog review
• AOB
  • AW webinar

Participants

• Ulf, Keith, Kevin Valdek, Ted, Michael Ger (Cloudera), Stephen Lawrence, Gunnar, Philippe

Minutes

Data Lake proposal

• Ulf: shows the following slide deck
• slide #2 - VSS data lake = VSS tree with data over time and VINs
  • Michael: will this be done dynamically ?
  • Ulf: let answer your question later
• slide #4 - VSS DB table structure
  
  • VIN table is the main entry key
  • how many tables will be generated ? actually quite a few table
  • at least more than 500 TV-nodes : 500 000 tables might cause a performance issue
  • this is why I changed the design to an alternative DB structure
• slide #5 - alternative DB structure
  
  • # of tables is proportional to the # VINs
  • might be a better design, need to get the opinion of a database expert
  • Michael: might get this reviewed by a DB expert
  • Keith: with one VIN table, performance wise it should be ok, storage wise of course this is different
• slide #6 - database definition
  • Ulf: we should not put the VSS path in the database, because if we change the VSS tree, we will have to spread the change in the DB
  • Ulf: a simplification of the tables is possible
  • discussion on the matches and data retrievals (not captured)
• slide #7 - workflow to retrieve data
  • Ted: I am wondering whether people wull use a traditional relational database, IMHO we need to get the opinion of Adnan and Benjamin on the database technology selection (i.e. in relationship with the graph server)
  • Gunnar; we are very early in the implementation, we are not at the point to reach any conclusion on the database technology to use
  • Philippe: mentions Sanjeev's comment on using object oriented data base
• slide #8 - deployment in CCS proof-of-concept
  • Ulf: I have already developed 95% of the VSStoDB adapter
• slide #9 - how to develop / publish this
  • Ulf: my intent is to small modifications in the VSS repo for storing the VSStoDB adapter code
• /TODO/ Ulf to ask opinion of Adnan and Benjamin on the VSS data lake design proposed in the presentation

Value measurement formats

• Kevin: review is still on my todo list, will provide feedback on Wednesday
• Ulf: same

Sprint & backlog review -  Jira tracker

• review of active sprint
• we start with the review of tickets assigned to Unknown User (kevinval)

  • Jira
    server JIRA serverId 121ddff2-c571-320f-9e4d-d5b9371533bd key CCS-81

    done, but the content of the wiki page is not deprecated, we might reopen the ticket in the future

  • Jira
    server JIRA serverId 121ddff2-c571-320f-9e4d-d5b9371533bd key CCS-8

    subtask 

    Jira
    server JIRA serverId 121ddff2-c571-320f-9e4d-d5b9371533bd key CCS-86

    is done, however parent task is not done since we need to create other subtasks on the definition of the proof-of-concept scope

  • Jira
    server JIRA serverId 121ddff2-c571-320f-9e4d-d5b9371533bd key CCS-96

    & 

    Jira
    server JIRA serverId 121ddff2-c571-320f-9e4d-d5b9371533bd key CCS-97

    created
• ticket assigned to Keith

  • Jira
    server JIRA serverId 121ddff2-c571-320f-9e4d-d5b9371533bd key CCS-88

    commented, for the time being we need a general understanding on the signal2service interface of Adaptive AUTOSAR
• other tickets will be reviewed in a later call

AOB - Automotive World Webinar

• Automotive World will schedule the webinar on 4 May, presenters will be Kevin and Gunnar, invite will be forward when it is available from AW

Wednesday,  April 1 - Communication Framework

...

• vehicle client
  • Philippe reports on the call with Sanjeev (Samsung) (on Monday this week)
  • Ulf: talked to Sanjeev, the two of them will team to work on the vehicle client
• in-vehicle OS
  • Philippe reports on the call with Bosch (on Tuesday this week) about the possible reuse of components from the Kuksa project
  • Sebastian Schildt:  Kuksa focused more on the in-vehicle stuff
  • Gunnar: this was our expectation
  • Sebastian cannot commit to deliver software to GENIVI with any deadline constraint, however he could offer
    
    • a Kuksa target board (1 or 2 samples) which is based on Rasperry Pi and can be plugged to the OBD
    • this OBD feeder could provide some sort of VSS feeder or a lower part of the VSS feeder that could be accessed through Web sockets
  • Tom said he is working on an infrastructure similar to CCS for the eBike project demo, there might be a cool alignment with CCS to look at. He will get back to us.
• detailed design
  • Gunnar: shows the UML diagrams he added to the CCS PoC wiki page, these diagrams represent the relationships between some of the components of the OEM cloud and Vehicle OS subsystems
  • Ulf: it is a good start, it makes things possible
  • Ulf: I am ready to talk about the SignalStore and specially with the Bosch people if they are ready to talk about it
  • Ulf: I thought about the data lake today, I will come with a proposal on next Monday likely, with a presentation, based on the use of a relational database like SQLite
  • Kevin V: the UML diagrams shown are for the vehicle interface, how do we describe things on the neutral server side ?
  • Gunnar: in my opinion the UML diagrams could be almost the same for the neutral server to OEM cloud interface,
  • Kevin V: the dependency would be the API definition of the OEM cloud (relates to the graphql schema)
  • discussion follows on the grapql schema content, how the VSS tree is represented with it ?
  • the team agree they need to develop their own understanding of a graphql schema
    • /TODO/ Kevin to take sample yaml code of VSS and transform it into a graphql schema
    • /TODO/ Philippe to ask Alex whether he can share the graphql he is using with Gunnar DONE
• next step
  • Philippe: expects the team to move now to the actual implementation of the first version of the PoC to be shown mid-May at the virtual technical summit

...