Deep Dive into Stateful Stream Processing in Structured Streaming - Tathagata Das

Поділитися
Вставка
  • Опубліковано 2 гру 2024

КОМЕНТАРІ • 10

  • @TE1gamingmadness
    @TE1gamingmadness 7 років тому +3

    We performed Streaming joins using Kafka Streams but faced a barrier with handling late data. Now, eagerly waiting for Spark 2.3 ;)

  • @takreem.akhter
    @takreem.akhter Рік тому

    14:21 with regards to deduplication, why not just use delta merge/upsert ?

  • @Dyslexic_Neuron
    @Dyslexic_Neuron 4 роки тому +1

    what if there is no data coming for any of the groups and wartermark doesnt progress . how will events get timedout in that case if we use eventtimetimeout ?

  • @nitishs3361
    @nitishs3361 4 роки тому

    Explained pretty well

  • @EldanSai
    @EldanSai 3 роки тому

    great talk

  • @megharaina6108
    @megharaina6108 3 роки тому

    Time 5:53 sec , speaker says maintaining the state thru checkpointing allows fault tolerance in both stateful and stateless streaming. As far as I understand we don't maintain the state in stateless streaming. How come stateless streaming become fault tolerant ?

    • @EldanSai
      @EldanSai 3 роки тому

      I think he might have meant that you'll still keep your offset saved through the checkpoint (assuming the source has some type of "offset") so in case of failure you'll continue from the last checkpoint.
      If you also have some state saved through aggregations, etc. Then as far as I understand, the "checkpoint" will save both the offset and the state (using WAL)

  • @Bowonfire
    @Bowonfire 3 роки тому +3

    What a weird way to make your speaker stop presenting at the end

  • @rexche9840
    @rexche9840 2 роки тому

    guys any tutorial on how i can stream crypto data from different sources?

  • @MyRania
    @MyRania 3 роки тому

    The guy at the end was NOT funny …