Real Time Data Processing

Slide 1

Slide 1 text

Inside KMK Real Time Data Processing Teguh Riadi

Slide 2

Slide 2 text

KMK DATA ARCHITECTURE

Slide 3

Slide 3 text

No content

Slide 4

Slide 4 text

Pub/Sub Plenty DATA INGESTION

Slide 5

Slide 5 text

Pub/Sub Plenty DATA INGESTION DATA STREAMING Dataflow

Slide 6

Slide 6 text

Pub/Sub Plenty DATA INGESTION DATA STREAMING Dataflow GCS Big Query DATA LAKE

Slide 7

Slide 7 text

Pub/Sub Plenty DATA INGESTION DATA STREAMING Dataflow GCS Big Query DATA LAKE

Slide 8

Slide 8 text

Pub/Sub Plenty DATA INGESTION DATA STREAMING Dataflow GCS Big Query DATA LAKE ETL DATA WAREHOUSE

Slide 9

Slide 9 text

Pub/Sub Plenty DATA INGESTION DATA STREAMING Dataflow GCS Big Query DATA LAKE ETL DATA WAREHOUSE VISUALIZATION Zeppelin

Slide 10

Slide 10 text

Pub/Sub Plenty DATA INGESTION DATA STREAMING Dataflow GCS Big Query DATA LAKE ETL DATA WAREHOUSE VISUALIZATION Zeppelin

Slide 11

Slide 11 text

KMK REAL TIME PROCESSING

Slide 12

Slide 12 text

LIVE STREAMING VIDEO

Slide 13

Slide 13 text

LIVE STREAMING VIDEO

Slide 14

Slide 14 text

LIVE STREAMING VIDEO NON LIVE STREAMING VIDEO

Slide 15

Slide 15 text

LIVE STREAMING VIDEO ON DEMAND

Slide 16

Slide 16 text

WHY WE NEED REAL TIME PROCESSING ?

Slide 17

Slide 17 text

Pub/Sub Plenty DATA INGESTION DATA STREAMING Dataflow Big Query DATA LAKE ETL DATA WAREHOUSE BATCH PROCESSING BEFORE USING REAL TIME PROCESSING

Slide 18

Slide 18 text

CASE : Vidio Alltime Count

Slide 19

Slide 19 text

DATA INGESTION DATA STREAMING DATA WAREHOUSE & LAKE BEFORE USING REAL TIME PROCESSING CASE : Vidio Alltime Count

Slide 20

Slide 20 text

DATA INGESTION DATA STREAMING DATA WAREHOUSE & LAKE ETL 2 ETL 1 ETL 3 BEFORE USING REAL TIME PROCESSING CASE : Vidio Alltime Count

Slide 21

Slide 21 text

DATA INGESTION DATA STREAMING DATA WAREHOUSE & LAKE ETL 2 ETL 1 ETL 3 ETL 4 BEFORE USING REAL TIME PROCESSING CASE : Vidio Alltime Count Every 5 min

Slide 22

Slide 22 text

DATA INGESTION DATA STREAMING DATA WAREHOUSE & LAKE ETL 2 ETL 1 ETL 3 ETL 4 BEFORE USING REAL TIME PROCESSING CASE : Vidio Alltime Count Every 5 min Not Fault Tolerance Not Real Time

Slide 23

Slide 23 text

Dataflow Pub/Sub REAL TIME PROCESSING

Slide 24

Slide 24 text

Pub/Sub Plenty DATA INGESTION DATA STREAMING Dataflow GCS Big Query DATA LAKE ETL DATA WAREHOUSE VISUALIZATION Zeppelin REAL TIME PROCESSING BATCH PROCESSING

Slide 25

Slide 25 text

REAL TIME PROCESSING CHALLENGE

Slide 26

Slide 26 text

CHALLENGE LIVE STREAMING & VIDEO COUNT VIEWER ACCURACY

Slide 27

Slide 27 text

CHALLENGE LIVE STREAMING & VIDEO COUNT VIEWER ACCURACY

Slide 28

Slide 28 text

CHALLENGE LIVE STREAMING & VIDEO COUNT VIEWER ACCURACY DATAFLOW SPIKE

Slide 29

Slide 29 text

CHALLENGE LIVE STREAMING & VIDEO COUNT VIEWER ACCURACY DATAFLOW SPIKE DATAFLOW SYSTEM LAG

Slide 30

Slide 30 text

CHALLENGE LIVE STREAMING & VIDEO COUNT VIEWER ACCURACY DATAFLOW SPIKE DATAFLOW SYSTEM LAG UN ACKNOWLEDGE MESSAGE

Slide 31

Slide 31 text

CHALLENGE LIVE STREAMING & VIDEO COUNT VIEWER ACCURACY DATAFLOW SPIKE DATAFLOW SYSTEM LAG UN ACKNOWLEDGE MESSAGE LATE DATA

Slide 32

Slide 32 text

CHALLENGE LIVE STREAMING & VIDEO COUNT VIEWER ACCURACY DATAFLOW SPIKE DATAFLOW SYSTEM LAG UN ACKNOWLEDGE MESSAGE LATE DATA

Slide 33

Slide 33 text

CHALLENGE LIVE STREAMING & VIDEO COUNT VIEWER ACCURACY GROWING DATA DATAFLOW SPIKE DATAFLOW SYSTEM LAG UN ACKNOWLEDGE MESSAGE LATE DATA

Slide 34

Slide 34 text

350 K Request / Minutes 2.5 M Request / Minutes Plenty GROWING DATA

Slide 35

Slide 35 text

350 K Request / Minutes 2.5 M Request / Minutes 127 K Events / Minutes 648 K Events / Minutes Plenty Live streaming GROWING DATA

Slide 36

Slide 36 text

350 K Request / Minutes 2.5 M Request / Minutes 127 K Events / Minutes 648 K Events / Minutes 368 Events / Minutes 3.2 K Events / Minutes Plenty Live streaming Video GROWING DATA * August 2018 Data

Slide 37

Slide 37 text

1 Request

Slide 38

Slide 38 text

Dataflow Spike

Slide 39

Slide 39 text

Dataflow System Lag

Slide 40

Slide 40 text

Un Acknowledge Message

Slide 41

Slide 41 text

Late Data

Slide 42

Slide 42 text

HOW WE TACKLE IT ?

Slide 43

Slide 43 text

ADD MORE DATAFLOW PIPELINE

Slide 44

Slide 44 text

ADD MORE DATAFLOW PIPELINE Dataflow Pub/Sub

Slide 45

Slide 45 text

ADD MORE DATAFLOW PIPELINE Dataflow Pub/Sub

Slide 46

Slide 46 text

ADD MORE DATAFLOW PIPELINE Dataflow Dataflow Pub/Sub Dataflow Pub/Sub Dataflow

Slide 47

Slide 47 text

ADD MORE DATAFLOW PIPELINE SPLIT PUBSUB TOPIC

Slide 48

Slide 48 text

SPLIT PUBSUB TOPIC Pub/Sub Ahoy.events

Slide 49

Slide 49 text

SPLIT PUBSUB TOPIC Pub/Sub Ahoy.events

Slide 50

Slide 50 text

SPLIT PUBSUB TOPIC Pub/Sub Ahoy.events Pub/Sub Ahoy.events Ahoy.events.livestreaming Ahoy.events.video.start

Slide 51

Slide 51 text

SPLIT PUBSUB TOPIC Pub/Sub Ahoy.events Pub/Sub Ahoy.events Ahoy.events.livestreaming Ahoy.events.video.start Dataflow Dataflow Dataflow Dataflow

Slide 52

Slide 52 text

ADD MORE DATAFLOW PIPELINE SPLIT PUBSUB EVENT ETL

Slide 53

Slide 53 text

Pub/Sub Plenty DATA INGESTION DATA STREAMING Dataflow REAL TIME PROCESSING

Slide 54

Slide 54 text

Pub/Sub Plenty DATA INGESTION DATA STREAMING Dataflow Big Query DATA LAKE ETL DATA WAREHOUSE REAL TIME PROCESSING BATCH PROCESSING

Slide 55

Slide 55 text

ADD MORE DATAFLOW PIPELINE SPLIT PUBSUB EVENT ETL MONITORING

Slide 56

Slide 56 text

MONITORING

Slide 57

Slide 57 text

MONITORING

Slide 58

Slide 58 text

MONITORING

Slide 59

Slide 59 text

THANK YOU

Slide 60

Slide 60 text

Pub/Sub Dataflow Big Query Batch Layer Speed Layer Serving Layer ETL 1 2 3 4 5