1.
Streaming Data Pipelines with
Apache Beam
Danny McCormick

2.
Agenda
● Who am I
● What is Apache Beam
● Beam Basics
● Processing streaming data
● Demo

3.
Who am I

4.
Me!

5.
What is Apache Beam

6.
In the beginning, there was MapReduce
Datastore
Map
Map
Map
Map
Map
Map
Reduce
Reduce
Reduce
Reduce
Reduce
Reduce
Shuffle
Datastore

7.
In the beginning, there was MapReduce

8.
Then came Flume (and Spark, Flink, and many more)
Datastore
Map
Map
Datastore
Map
Group by Key
(Reduce)
Combine
Map
Map
Combine
Datastore
Datastore
Datastore

9.
From Flume came Beam
Datastore
Map
Map
Datastore
Map
Group by Key
(Reduce)
Combine
Map
Map
Combine
Datastore
Datastore
Datastore

10.
Unified Model for Batch and Streaming
● Batch processing is a special case of
stream processing
● Batch + Stream = Beam

11.
Build your pipeline in whatever language(s) you want…
Group by Key

12.
… with whatever execution engine you want
Cloud Dataflow
Apache Spark
Apache Flink
Apache Apex
Gearpump
Apache Samza
Apache Nemo
(incubating)
IBM Streams
Group by Key

13.
Beam Basics

14.
Terms
● PCollection - distributed multi-element
dataset
● Transform - operation that takes N
PCollections and produces M PCollections
● Pipeline - directed acyclic graph of
Transforms and PCollections

15.
Basic Beam Graph
Source
Transform
Sink
Transform
Source
Transform
Map
Transform
Combine
Transform
Sink
Transform
Sink
Transform

16.
Basic Beam Pipeline
def add_one(element):
return element + 1
import apache_beam as beam
with beam.Pipeline() as pipeline:
pipeline
| beam.io.ReadFromText('gs://some/inputData.txt')
| beam.Map(add_one)
| beam.io.WriteToText('gs://some/outputData')
Read Text
file
Map
Transform
Write to text
file

17.
How to use Beam to process huge
amounts of streaming data

18.
We want to go from this:

19.
To this:
Monday
Tuesday
Wednesday
Thursday
Friday

20.
To this:
9:00
8:00 14:00
13:00
12:00
11:00
10:00

21.
Streaming data might be:
● Late
● Incomplete
● Rate limited
● Infinite

22.
You will need to make tradeoffs between:
● Cost
● Completeness
● Low Latency

23.
Example 1: Billing Pipeline
Completeness Low Latency Low Cost
Important
Not Important

24.
Example 2: Billing Estimator
Completeness Low Latency Low Cost
`
Important
Not Important

25.
Example 3: Fraud Detection
Completeness Low Latency Low Cost
`
Important
Not Important

26.
Windows
9:00
8:00 14:00
13:00
12:00
11:00
10:00
Aggregate or output Aggregate or output Aggregate or output Aggregate or outpu
output

27.
Fixed Windows
9:00
8:00 14:00
13:00
12:00
11:00
10:00
Aggregate
or output
Aggregate
or output
Aggregate
or output
Aggregate
or output
Aggregate
or output
Aggregate
or output
Aggregate
or output
ggregate
r output

28.
Sliding Windows
Aggregate or output Aggregate or output Aggregate or output
Aggregate
or
output Aggregate
or
output Aggregate
or
output output
Agg
Aggregate or output

29.
Sliding Windows
9:00
8:00 14:00
13:00
12:00
11:00
10:00
Aggregate or output
Aggregate or output
ate or output
Aggregate or o
Aggregate or output
Aggregate or output
Aggregate or output
Aggregate or output

30.
Session Windows
9:00
8:00 14:00
13:00
12:00
11:00
10:00
Aggregate or
output
Aggregate or output
Aggregate or
output
Aggregate or output
A

31.
Global Window
9:00
8:00 14:00
13:00
12:00
11:00
10:00

32.
Code
● items | beam.WindowInto(window.FixedWindows(60)) # 60s fixed windows
● items | beam.WindowInto(window.SlidingWindows(30, 5)) # 30s sliding window every 5s
● items | beam.WindowInto(window.Sessions(10 * 60)) # window breaks after 10 empty min
● items | beam.WindowInto(window.GlobalWindows()) # single global window

33.
Real Time vs Event Time - Expectation
Event Time
Processing
Time

34.
Real Time vs Event Time - Reality
Event Time
Processing
Time

35.
How do we know its safe to finish a window’s work?
Event Time
Processing
Time

36.
Processing Time?
Event Time
Processing
Time

37.
Processing Time? Lots of late data won’t be counted
Event Time
Processing
Time

38.
Beam’s Solution - Watermarks!
Event Time
Processing
Time

39.
Watermarks
● Beam’s notion of when data is
complete
● When a watermark passes the end of
a window, additional data is late
● Beam has several built in watermark
estimators

40.
Example: Timestamp observing estimation
Event Time
Processing
Time

41.
Example: Timestamp observing estimation
Event Time
Processing
Time

42.
Example: Timestamp observing estimation
Event Time
Processing
Time

43.
Example: Timestamp observing estimation
Event Time
Processing
Time

44.
Example: Timestamp observing estimation
Event Time
Processing
Time

45.
Example: Timestamp observing estimation
Event Time
Processing
Time

46.
Example: Timestamp observing estimation
Event Time
Processing
Time

47.
Example: Timestamp observing estimation
Event Time
Processing
Time

48.
Example: Timestamp observing estimation
Event Time
Processing
Time

49.
Example: Timestamp observing estimation
Event Time
Processing
Time

50.
Example: Timestamp observing estimation
Event Time
Processing
Time

51.
Example: Timestamp observing estimation
Event Time
Processing
Time

52.
Example: Timestamp observing estimation
Event Time
Processing
Time
Late Data*

53.
Watermarks
● Handled at the source I/O level
● Most pipelines don’t need to
implement estimation, but do need to
be aware of it

54.
Recall Tradeoffs
Completeness Low Latency Low Cost
`
Important
Not Important

55.
Triggers
● Beam’s mechanism for controlling
tradeoffs
● Describe when to emit aggregated
results of a single window
● Allow emitting early results or results
including late data

56.
Types of Triggers
● Event Time Triggers
● Processing Time Triggers
● Data-Driven Triggers
● Composite Triggers

57.
Set on windows
pcollection | WindowInto(
FixedWindows(1 * 60),
trigger=AfterProcessingTime(1 * 60),
accumulation_mode=AccumulationMode.DISCARDING)

58.
Example Triggers
● AfterProcessingTime(delay=1 * 60)
● AfterCount(1)
● AfterWatermark(
early=AfterProcessingTime(delay=1 * 60),
late=AfterCount(1))
● AfterAny(AfterCount(1),
AfterProcessingTime(delay=1 * 60))

59.
Accumulation Mode
● Describes how to handle data that
has already been emitted
● 2 types: Accumulating and
Discarding

60.
Discarding Accumulation Mode
pcollection | WindowInto(
FixedWindows(1 * 60),
trigger=Repeating(AfterCount(3)),
accumulation_mode=AccumulationMode.DISCARDING)
[5, 8, 3, 1, 2, 6, 9, 7]

61.
Discarding Accumulation Mode
pcollection | WindowInto(
FixedWindows(1 * 60),
trigger=Repeating(AfterCount(3)),
accumulation_mode=AccumulationMode.DISCARDING)
[5, 8, 3, 1, 2, 6, 9, 7] -> [5, 8, 3]

62.
Discarding Accumulation Mode
pcollection | WindowInto(
FixedWindows(1 * 60),
trigger=Repeating(AfterCount(3)),
accumulation_mode=AccumulationMode.DISCARDING)
[5, 8, 3, 1, 2, 6, 9, 7] -> [5, 8, 3]
[1, 2, 6]

63.
Discarding Accumulation Mode
pcollection | WindowInto(
FixedWindows(1 * 60),
trigger=Repeating(AfterCount(3)),
accumulation_mode=AccumulationMode.DISCARDING)
[5, 8, 3, 1, 2, 6, 9, 7] -> [5, 8, 3]
[1, 2, 6]
[9, 7]

64.
Discarding Accumulation Mode
pcollection | WindowInto(
FixedWindows(1 * 60),
trigger=Repeating(AfterCount(3)),
accumulation_mode=AccumulationMode.Accumulating)
[5, 8, 3, 1, 2, 6, 9, 7]

65.
Discarding Accumulation Mode
pcollection | WindowInto(
FixedWindows(1 * 60),
trigger=Repeating(AfterCount(3)),
accumulation_mode=AccumulationMode.Accumulating)
[5, 8, 3, 1, 2, 6, 9, 7] -> [5, 8, 3]

66.
Discarding Accumulation Mode
pcollection | WindowInto(
FixedWindows(1 * 60),
trigger=Repeating(AfterCount(3)),
accumulation_mode=AccumulationMode.Accumulating)
[5, 8, 3, 1, 2, 6, 9, 7] -> [5, 8, 3]
[5, 8, 3, 1, 2, 6]

67.
Discarding Accumulation Mode
pcollection | WindowInto(
FixedWindows(1 * 60),
trigger=Repeating(AfterCount(3)),
accumulation_mode=AccumulationMode.Accumulating)
[5, 8, 3, 1, 2, 6, 9, 7] -> [5, 8, 3]
[5, 8, 3, 1, 2, 6]
[5, 8, 3, 1, 2, 6, 9, 7]

68.
More!
● Pipeline State
● Timers
● Runner initiated splits
● Self checkpointing
● Bundle finalization

69.
Demo
https://github.com/damccorm/ato-demo-2022

70.
Come join our community!

71.
Questions?
Slides - shorturl.at/GNU07