LG Electronics CTO Division: Workload orchestration between vehicle and cloud using DMS

00:11:19
https://www.youtube.com/watch?v=ywMa71XPjec

Resumen

TLDRIn deze presentatie geeft DK Zhang van LG Electronics een overzicht van de orkestratie van workloads tussen voertuigen en de cloud in gemengde kritieke omgevingen. Het focus ligt op het orkestratieplatform genaamd Piccolo, wat een abstractielaag biedt voor de integratie van voertuigscenario's. LG streeft naar efficiënte softwarebeheer door cloud-native benaderingen en microservice architectuur (MSA) in Software Defined Vehicles (SDV). Zhang bespreekt de voordelen van Piccolo in het bieden van een stabiele architectuur, efficiënte middelenbeheer en eenvoudige schaling van autoservices. Een proof-of-concept (POC) laat automatische opschaling en veiligheidsfunctie-integratie zien via Piccolo, terwijl uitdagingen in de transitie van monolithische naar microservice architecturen worden besproken. Uiteindelijk benadrukt het de noodzaak van MSA in de automobielindustrie om veiligheids- en tijdskritieke uitdagingen effectief aan te pakken.

Para llevar

  • 👨‍💻 DK Zhang van LG presenteert over orkestratie tussen voertuigen en de cloud.
  • 🚗 Focus op SDV, software defined vehicle platformen.
  • 🛠️ Piccolo platform biedt abstractie voor voertuigscenario-integratie.
  • ☁️ Cloud-native benaderingen verbeteren softwarebeheer.
  • 🧩 Overgang van monolithische naar microservice-architecturen is uitdagend.
  • 🔧 MSA biedt voordelen in ontwikkeling en integratie.
  • 🔍 Piccolo zorgt voor stabiliteit en efficiënt middelenbeheer.
  • 📈 Automatische opschaling getoond in POC.
  • 🔐 Veiligheidsfunctie-integratie benadrukt in demo.
  • 🌐 Belang van standaardisatie en samenwerking in orkestratie.

Cronología

  • 00:00:00 - 00:05:00

    DK Zhang van LG Electronics presenteert over hoe ze werken aan de orkestratie van workloads tussen voertuigen en de cloud in een omgeving met gemengde kritiek. Hij bespreekt de noodzaak van een orkestratiesysteem in voertuigen, aangedreven door microservice-architectuur, om de efficiëntie en herbruikbaarheid van software te maximaliseren. LG heeft een systeem ontwikkeld genaamd Piccolo dat een abstractie-API biedt voor de integratie van voertuigscenario's binnen een STV (softwaregedefinieerd voertuig) omgeving. Deze oplossing richt zich op het verbeteren van de veiligheid en beveiliging binnen een cloudgebaseerde benadering, door het decomponeren van monolithische diensten naar microservices.

  • 00:05:00 - 00:11:19

    Piccolo's POC-concept laat zien hoe containerorkestratie kan bijdragen aan autoschaling en de toevoeging van nieuwe veiligheidsfuncties. Bij autoschaling kan het systeem automatisch uitbreiden naar verbonden apparaten als de gewenste prestaties niet worden bereikt. Voor veiligheidsfuncties kunnen er extra modules worden toegevoegd, zoals bij het omschakelen van rijmodi, wat kan leiden tot nieuwe veiligheidsfuncties indien nodig. Deze capaciteiten tonen aan hoe Piccolo kan zorgen voor effectieve resourcebeheer en samenwerking tussen voertuig en cloud, zelfs met heterogene kritische eisen. Het systeem is gebouwd op het Ava-platform en maakt gebruik van Amazon-cloudfunctionaliteiten.

Mapa mental

Mind Map

Preguntas frecuentes

  • Wie is de spreker?

    De spreker is DK Zhang, een technisch fellow van LG Electronics.

  • Wat is het hoofdonderwerp van deze presentatie?

    Het hoofdonderwerp is workload orkestratie tussen voertuigen en de cloud in een gemengde kritieke omgeving.

  • Wat is Piccolo?

    Piccolo is een orkestratieplatform ontwikkeld door LG voor Software Defined Vehicle (SDV) omgevingen.

  • Waarom is orkestratie belangrijk in automobielsoftware?

    Orkestratie is belangrijk om efficiënte softwarebeheer te bieden en de herbruikbaarheid van software te maximaliseren in gemengde kritieke omgevingen.

  • Wat zijn enkele uitdagingen bij het gebruik van MSA in voertuigen?

    Uitdagingen omvatten de decompositie van monolithische services met kritieke veiligheids- en tijdseisen en de bemoeilijking van softwarebeheer door complexiteit.

  • Hoe verbetert MSA de automobielomgeving?

    MSA biedt voordelen zoals snellere ontwikkeling, integratie en implementatie van software, zelfs na de lancering van een voertuig.

  • Wat is een POC-concept?

    Een POC-concept (Proof of Concept) toont de haalbaarheid van orkestratiegebruik door middel van twee gebruiksgevallen: automatische opschaling en toevoeging van nieuwe veiligheidsfuncties.

  • Wat is de rol van veiligheid en tijdkritieke criteria in MSA?

    Ze zijn aanzienlijke criteria voor het decomponeren van monolithische services in microservices om kritieke functies veilig te beheren.

  • Wat toont de demo in de presentatie aan?

    De demo laat de automatische schaalbaarheid en toegevoegde veiligheidsfuncties zien in een voertuigsysteem door gebruik van Piccolo.

Ver más resúmenes de vídeos

Obtén acceso instantáneo a resúmenes gratuitos de vídeos de YouTube gracias a la IA.
Subtítulos
en
Desplazamiento automático:
  • 00:00:00
    all right all right uh good morning good
  • 00:00:03
    evening everyone
  • 00:00:04
    I appreciate uh post office to give a
  • 00:00:08
    chance to have a presentation today uh
  • 00:00:12
    nice to meet you or something member in
  • 00:00:13
    here uh let me briefly introduce myself
  • 00:00:16
    I'm DK Zhang a technical fellow from LG
  • 00:00:20
    electronic I'm belonging to smart mobile
  • 00:00:23
    lab under the LGE CTO organization
  • 00:00:26
    uh we are focusing
  • 00:00:29
    on the best automotive are in the
  • 00:00:31
    specifically STV software defined
  • 00:00:33
    vehicle platform development
  • 00:00:36
    but I'm going to talk about today is
  • 00:00:39
    workload orchestration between a vehicle
  • 00:00:41
    and cloud in a mixed critical
  • 00:00:43
    environment
  • 00:00:44
    the presentation will focus on how to
  • 00:00:47
    achieve the mixed critical orchestration
  • 00:00:49
    and also LGE orchestration operator name
  • 00:00:54
    is Piccolo
  • 00:00:57
    so okay uh most important question is
  • 00:01:01
    how to achieve the mixed critical
  • 00:01:03
    orchestration for sdb
  • 00:01:05
    people we move on specific MCO we pretty
  • 00:01:08
    touch why STV and what is OEM pain point
  • 00:01:11
    and also what is the customer value
  • 00:01:14
    Automotive typically is the same
  • 00:01:16
    critical and especially Hardware time
  • 00:01:19
    requirement and many bigger service
  • 00:01:21
    exist in mixed critical environment
  • 00:01:24
    and also most of the Legacy service is
  • 00:01:27
    still monolithic approach not micro
  • 00:01:29
    service so far
  • 00:01:31
    on the other hand
  • 00:01:32
    MCO will have a good way to enhancing
  • 00:01:35
    Safety and Security based on cloud
  • 00:01:38
    native approach and MSA for sdb and also
  • 00:01:42
    provide the efficient software
  • 00:01:43
    management to achieve the maximization
  • 00:01:46
    of the Softail reusability and service
  • 00:01:49
    Etc
  • 00:01:50
    just by doing that the continuous app
  • 00:01:53
    and service can be provided thanks to a
  • 00:01:55
    microservice architecture
  • 00:01:58
    with that in mind there is still no de
  • 00:02:01
    facto standard regarding the
  • 00:02:03
    orchestration specifically vehicle
  • 00:02:05
    interface
  • 00:02:06
    that's the reason why we developed the
  • 00:02:08
    piccolo this is orchestration
  • 00:02:10
    abstraction called sdb
  • 00:02:12
    specifically provide the abstraction API
  • 00:02:16
    for integration of the vehicle scenario
  • 00:02:18
    and we will have a vision to achieve the
  • 00:02:21
    stable architecture against the two
  • 00:02:23
    mixed critical environment at the end of
  • 00:02:27
    the day Piccolo can make sure that
  • 00:02:30
    effective Resource Management with the
  • 00:02:31
    criticality and also easy to manage a
  • 00:02:34
    bigger service using contain and
  • 00:02:36
    infrastructure
  • 00:02:37
    finally we can reduce development timing
  • 00:02:40
    and cost
  • 00:02:44
    so this is microservice architecture on
  • 00:02:48
    explanation
  • 00:02:50
    you know automotively any architecture
  • 00:02:52
    are evolving into MSA as I mentioned
  • 00:02:54
    earlier specifically for sdb environment
  • 00:02:57
    in a sense of the regressive monolithic
  • 00:02:59
    Beaker service are integrated critical
  • 00:03:01
    and non-critical feature so partial
  • 00:03:04
    software update is not realistic and
  • 00:03:07
    also software management is very tricky
  • 00:03:09
    due to the software complexity even
  • 00:03:12
    mixed critical environment
  • 00:03:14
    but you know MSA can solve this kind of
  • 00:03:17
    issue MSA has a lot of advantage of the
  • 00:03:20
    First Development integration and
  • 00:03:22
    development deployment
  • 00:03:24
    even if the car has been launched
  • 00:03:27
    at the end of the day isn't necessary to
  • 00:03:29
    create the MSA for sdb why considering
  • 00:03:33
    the mixed critical environment
  • 00:03:35
    when it comes to Ms it is necessary to
  • 00:03:37
    consider critical it has important
  • 00:03:39
    criteria for decomposing Legacy
  • 00:03:41
    monolithic servicing to microservice
  • 00:03:45
    at the moment the safety and time
  • 00:03:47
    criticality award important criteria
  • 00:03:50
    while decomposition of the beaker
  • 00:03:53
    service to create MSA in a vehicle
  • 00:03:56
    when operating to comfort the bigger
  • 00:03:58
    service is support important to
  • 00:04:00
    determine the service location and the
  • 00:04:03
    data test between the according to
  • 00:04:06
    criticalities such as the safety
  • 00:04:08
    real-time or non-critical
  • 00:04:10
    however most big Chinese parties the
  • 00:04:12
    real-time criteria in compared to
  • 00:04:15
    monolithic adding communication protocol
  • 00:04:18
    can bring in additional latency and
  • 00:04:21
    performance issue
  • 00:04:23
    here we have a DMS example DNS means the
  • 00:04:26
    driving monitor system to explain how to
  • 00:04:28
    break it down to microservice by
  • 00:04:30
    considering safety feature and real-time
  • 00:04:33
    components
  • 00:04:37
    so this is Piccolo so here we specify
  • 00:04:39
    the key color requirement to achieve the
  • 00:04:41
    sdb platform there are various type of
  • 00:04:43
    computing resources and service
  • 00:04:45
    distribution
  • 00:04:47
    orchestration abstraction layer based on
  • 00:04:50
    the container infrastructure to what is
  • 00:04:52
    providing bigger orchestration API and
  • 00:04:56
    declarative configuration
  • 00:04:58
    also needed to monitor orchestration
  • 00:05:00
    action what's going on in terms of not
  • 00:05:03
    part and service Etc
  • 00:05:05
    more specifically needed to Define what
  • 00:05:08
    kind of law associate with each node
  • 00:05:10
    such as vehicle Edge and Cloud
  • 00:05:14
    if you look at the glow table here you
  • 00:05:17
    can see the load of the vehicle is
  • 00:05:19
    control bigger service and the other
  • 00:05:21
    load for Edge cloud is Computing
  • 00:05:23
    resource and also bigger size should
  • 00:05:26
    consider safety and real time
  • 00:05:29
    on the other hand the edge Cloud side
  • 00:05:32
    let's say between our New York time
  • 00:05:34
    feature can be considered
  • 00:05:36
    this part I expect to make the
  • 00:05:38
    collaboration in Sophie such as making
  • 00:05:41
    some standardization
  • 00:05:43
    we believe that the value of a piccolo
  • 00:05:45
    is is to integrate the bigger specific
  • 00:05:47
    service suppose to stable architecture
  • 00:05:50
    and make sure that enhancing the some
  • 00:05:52
    sort of resource utilization
  • 00:05:57
    so this is a POC concept so we are going
  • 00:06:00
    to show two use case about the container
  • 00:06:03
    of castration just by using the piccolo
  • 00:06:06
    the one is the auto scaling
  • 00:06:09
    uh
  • 00:06:10
    based on the Apple PS so shenario is
  • 00:06:12
    like this so when do not reach the
  • 00:06:15
    desired performance
  • 00:06:17
    or when Computing resource exists that
  • 00:06:19
    this means that the resource is not
  • 00:06:21
    available some some sort of establish
  • 00:06:24
    and happening so at this moment that
  • 00:06:26
    automatically just the scale out to uh
  • 00:06:29
    some relatively connected device
  • 00:06:32
    so that's the one uh scenario the second
  • 00:06:36
    scenario is that the adding new safety
  • 00:06:39
    MSA picture
  • 00:06:40
    the scenario is like uh some driving
  • 00:06:43
    mode change so if you look at the below
  • 00:06:46
    uh
  • 00:06:47
    little picture the parking mode parking
  • 00:06:50
    mode is a basically no required for the
  • 00:06:53
    sum
  • 00:06:54
    70 pitcher but
  • 00:06:56
    driving mode uh DMS case uh probably
  • 00:07:00
    some sort of a eye detection is required
  • 00:07:02
    that is actually related to the some
  • 00:07:06
    functional safety because people
  • 00:07:10
    driver maybe some drowsiness or sleep in
  • 00:07:14
    that sense that the accident may happen
  • 00:07:17
    so in this sense that we just uh
  • 00:07:22
    uploading some Facebook picture that's
  • 00:07:26
    not actually not uh critical portion
  • 00:07:29
    then we're just uploading to the to the
  • 00:07:31
    cloud just by doing that we can make
  • 00:07:33
    some workload orchestration uh Behavior
  • 00:07:38
    so here is the PLC system architecture
  • 00:07:41
    that we designed so we are using the uh
  • 00:07:44
    the Ava platform by adelink so the we
  • 00:07:49
    implementing just the EOL Hill and on
  • 00:07:53
    top of that we implementing Piccolo or
  • 00:07:56
    LG operator
  • 00:07:57
    and this is the masternode and also we
  • 00:08:01
    have some CDC uh this is going to be the
  • 00:08:03
    worker note and also we are using the
  • 00:08:05
    the Amazon uh correlation just for the
  • 00:08:08
    cloud so this is the the big picture for
  • 00:08:11
    the our pocs system architecture
  • 00:08:16
    all right so the we are going to show
  • 00:08:18
    you the demo
  • 00:08:20
    sorry
  • 00:08:23
    the first
  • 00:08:25
    demo is automatically just the scale
  • 00:08:29
    that you can see that the left picture
  • 00:08:32
    is the DMS use case
  • 00:08:34
    in the beginning so you can see the
  • 00:08:36
    lower the FPS prime rate is pretty low
  • 00:08:40
    but we if we just uh
  • 00:08:43
    some sort of assumption is basically
  • 00:08:45
    just the starvation the resource is not
  • 00:08:47
    available
  • 00:08:48
    at this moment that uh automatically
  • 00:08:51
    the scale out to the relatively
  • 00:08:55
    connected device in here we are using
  • 00:08:57
    the CDC device
  • 00:08:58
    so you can see that the uh the prime net
  • 00:09:01
    prime rate is going up or higher so at
  • 00:09:05
    the beginning around the time but it
  • 00:09:08
    goes up to around 20 something
  • 00:09:11
    so you can see that the perpetually just
  • 00:09:13
    walk through the orchestration is
  • 00:09:16
    happening
  • 00:09:19
    okay the
  • 00:09:22
    the next demo
  • 00:09:26
    next demo is that the adding new save
  • 00:09:29
    dmsa feature as I mentioned before our
  • 00:09:32
    assumption is that the
  • 00:09:34
    uh some driving mode change so in the
  • 00:09:37
    beginning there is no required some uh
  • 00:09:40
    I'll
  • 00:09:41
    detection function so
  • 00:09:44
    but if we change the subdriving mode I
  • 00:09:47
    detection may happen so this can be
  • 00:09:51
    activated some sort of some
  • 00:09:54
    for functional safety so
  • 00:09:58
    at this moment that we can make uh
  • 00:10:01
    create the multiple part
  • 00:10:03
    uh to
  • 00:10:06
    something like some uploading to the
  • 00:10:10
    uh not only does LG device but also
  • 00:10:13
    Cloud Amazon Cloud so here you can see
  • 00:10:16
    that the
  • 00:10:18
    sorry
  • 00:10:21
    sorry here you can see that cloud side
  • 00:10:24
    that we are
  • 00:10:26
    uploading something like a
  • 00:10:28
    something like a non uh functional
  • 00:10:33
    safety feature we just uploading here
  • 00:10:35
    but in the beaker side that we assigned
  • 00:10:37
    the High detection
  • 00:10:40
    mostly related to the function of safety
  • 00:10:42
    those individual
  • 00:10:44
    so
  • 00:10:48
    simply say that the workload the
  • 00:10:50
    workload the orchestration is happening
  • 00:10:53
    even though the adding new pitch just by
  • 00:10:56
    using the Microsoft
  • 00:10:57
    oh this is the end of my presentation
  • 00:11:01
    thank you for listening and if you have
  • 00:11:04
    any further question uh please let me
  • 00:11:07
    know please just using this email and we
  • 00:11:10
    will get back to you with us
  • 00:11:12
    thank you very much
  • 00:11:14
    thank you very much DK
  • 00:11:17
    thank you
Etiquetas
  • Piccolo
  • orchestratie
  • automobielindustrie
  • veiligheid
  • microservices
  • SDV
  • cloud-native
  • softwarebeheer
  • LG Electronics
  • innovatie