Introducing Kafka's Streams API

1Confidential
Introducing Kafka’s Streams API
Stream processing made simple
Target audience: technical staff, developers, architects
Expected duration for full deck: 45 minutes

2Confidential
0.10 Data processing (Streams API)
0.9 Data integration (Connect API)
Intra-cluster
replication
0.8
Apache Kafka: birthed as a messaging system, now a streaming platform
2012 2014 2015 2016 2017
Cluster mirroring,
data compression
0.7
2013

3Confidential
Kafka’s Streams API: the easiest way to process data in Apache Kafka
Key Benefits of Apache Kafka’s Streams API
• Build Apps, Not Clusters: no additional cluster required
• Cluster to go: elastic, scalable, distributed, fault-tolerant, secure
• Database to go: tables, local state, interactive queries
• Equally viable for S / M / L / XL / XXL use cases
• “Runs Everywhere”: integrates with your existing deployment
strategies such as containers, automation, cloud
Part of open source Apache Kafka, introduced in 0.10+
• Powerful client library to build stream processing apps
• Apps are standard Java applications that run on client
machines
• https://github.com/apache/kafka/tree/trunk/streams
Streams
API
Your App
Kafka
Cluster

4Confidential
Kafka’s Streams API: Unix analogy
$ cat < in.txt | grep “apache” | tr a-‐z A-‐Z > out.txt
Kafka Cluster
Connect API Streams API

5Confidential
Streams API in the context of Kafka
Streams
API
Your App
Kafka
Cluster
ConnectAPI
ConnectAPI
OtherSystems
OtherSystems

6Confidential
When to use Kafka’s Streams API
• Mainstream Application Development
• To build core business applications
• Microservices
• Fast Data apps for small and big data
• Reactive applications
• Continuous queries and transformations
• Event-triggered processes
• The “T” in ETL
• <and more>
Use case examples
• Real-time monitoring and intelligence
• Customer 360-degree view
• Fraud detection
• Location-based marketing
• Fleet management
• <and more>

7Confidential
Some public use cases in the wild & external articles
• Applying Kafka’s Streams API for internal message delivery pipeline at LINE Corp.
• http://developers.linecorp.com/blog/?p=3960
• Kafka Streams in production at LINE, a social platform based in Japan with 220+ million users
• Microservices and reactive applications at Capital One
• https://speakerdeck.com/bobbycalderwood/commander-decoupled-immutable-rest-apis-with-kafka-streams
• User behavior analysis
• https://timothyrenner.github.io/engineering/2016/08/11/kafka-streams-not-looking-at-facebook.html
• Containerized Kafka Streams applications in Scala
• https://www.madewithtea.com/processing-tweets-with-kafka-streams.html
• Geo-spatial data analysis
• http://www.infolace.com/blog/2016/07/14/simple-spatial-windowing-with-kafka-streams/
• Language classification with machine learning
• https://dzone.com/articles/machine-learning-with-kafka-streams

8Confidential
Do more with less

9Confidential
Architecture comparison: use case example
Real-time dashboard for security monitoring
“Which of my data centers are under attack?”

10Confidential
Architecture comparison: use case example
Other
App
Dashboard
Frontend
App
Other
App
1 Capture business
events in Kafka
2 Must process events with
separate cluster (e.g. Spark)
4
Other apps access latest results
by querying these DBs
3 Must share latest results through
separate systems (e.g. MySQL)
Before: Undue complexity, heavy footprint, many technologies, split ownership with conflicting
priorities
Your
“Job”
Other
App
Dashboard
Frontend
App
Other
App
1 Capture business
events in Kafka
2 Process events with standard
Java apps that use Kafka Streams
3 Now other apps can directly
query the latest results
With Kafka Streams: simplified, app-centric architecture, puts app owners in control
Kafka
Streams
Your App

13Confidential
How do I install the Streams API?
• There is and there should be no “installation” – Build Apps, Not Clusters!
• It’s a library. Add it to your app like any other library.
<dependency>
<groupId>org.apache.kafka</groupId>
<artifactId>kafka-‐streams</artifactId>
<version>0.10.1.1</version>
</dependency>

14Confidential
“But wait a minute – where’s THE CLUSTER to process the data?”
• No cluster needed – Build Apps, Not Clusters!
• Unlearn bad habits: “do cool stuff with data ≠ must have cluster”
Ok. Ok. Ok.

15Confidential
Organizational benefits: decouple teams and roadmaps, scale people

16Confidential
Organizational benefits: decouple teams and roadmaps, scale people
Infrastructure Team
(Kafka as a shared, multi-tenant service)
Fraud
detection
app
Payments team
Recommenda
tions app
Mobile team
Security
alerts
app
Operations team
...more apps...
...

17Confidential
How do I package, deploy, monitor my apps? How do I …?
• Whatever works for you. Stick to what you/your company think is the best way.
• No magic needed.
• Why? Because an app that uses the Streams API is…a normal Java app.

19Confidential
The API is but the tip of the iceberg
API, coding
Org. processes
Reality™
Deployment
Operations
Security
…
Architecture
Debugging

20Confidential
• API option 1: DSL (declarative)
KStream<Integer, Integer> input =
builder.stream("numbers-‐topic");
// Stateless computation
KStream<Integer, Integer> doubled =
input.mapValues(v -‐> v * 2);
// Stateful computation
KTable<Integer, Integer> sumOfOdds = input
.filter((k,v) -‐> v % 2 != 0)
.selectKey((k, v) -‐> 1)
.groupByKey()
.reduce((v1, v2) -‐> v1 + v2, "sum-‐of-‐odds");
The preferred API for most use cases.
Particularly appeals to:
• Fans of Scala, functional programming
• Users familiar with e.g. Spark

21Confidential
• API option 2: Processor API (imperative)
class PrintToConsoleProcessor
implements Processor<K, V> {
@Override
public void init(ProcessorContext context) {}
@Override
void process(K key, V value) {
System.out.println("Got value " + value);
}
@Override
void punctuate(long timestamp) {}
@Override
void close() {}
}
Full flexibility but more manual work
Appeals to:
• Users who require functionality that is
not yet available in the DSL
• Users familiar with e.g. Storm, Samza
• Still, check out the DSL!

22Confidential
When to use Kafka Streams vs. Kafka’s “normal” consumer clients
Kafka Streams
• Basically all the time
Kafka consumer clients (Java, C/C++, Python, Go, …)
• When you must interact with Kafka at a very low
level and/or in a very special way
• Example: When integrating your own stream
processing tool (Spark, Storm) with Kafka.

23Confidential
Code comparison
Featuring Kafka with Streams API <-> Spark Streaming

24Confidential
”My WordCount is better than your WordCount” (?)
Kafka
Spark
These isolated code snippets are nice (and actually quite similar) but they are not very meaningful. In practice, we
also need to read data from somewhere, write data back to somewhere, etc.– but we can see none of this here.

25Confidential
WordCount in Kafka
Word
Count

26Confidential
Compared to: WordCount in Spark 2.0
1
2
3
Runtime model leaks into
processing logic
(here: interfacing from
Spark with Kafka)

27Confidential
Compared to: WordCount in Spark 2.0
4
5
Runtime model leaks into
processing logic
(driver vs. executors)

31Confidential
Key concepts
Kafka Core Kafka Streams

32Confidential
Streams and Tables
Stream Processing meets Databases

35Confidential
Key observation: close relationship between Streams and Tables
http://www.confluent.io/blog/introducing-kafka-streams-stream-processing-made-simple
http://docs.confluent.io/current/streams/concepts.html#duality-of-streams-and-tables

37Confidential
Example: Streams and Tables in Kafka
Word Count
hello 2
kafka 1
world 1
… …

42Confidential
Example: continuously compute current users per geo-region
4
7
5
3
2
8 4
7
6
3
2
7
Alice
Real-time dashboard
“How many users younger than 30y, per region?”
alice Europe
user-locations
alice Asia, 25y, …
bob Europe, 46y, …
… …
alice Europe, 25y, …
… …
-1
+1
user-locations
(mobile team)
user-prefs
(web team)

43Confidential
KTable<UserId, Location> userLocations = builder.table(“user-‐locations-‐topic”);
KTable<UserId, Prefs> userPrefs = builder.table(“user-‐preferences-‐topic”);

44Confidential
alice Europe
user-locations
… …
… …
// Merge into detailed user profiles (continuously updated)
KTable<UserId, UserProfile> userProfiles =
userLocations.join(userPrefs, (loc, prefs) -‐> new UserProfile(loc, prefs));
KTable userProfilesKTable userProfiles

45Confidential
// Compute per-‐region statistics (continuously updated)
KTable<UserId, Long> usersPerRegion = userProfiles
.filter((userId, profile) -‐> profile.age < 30)
.groupBy((userId, profile) -‐> profile.location)
.count();
alice Europe
user-locations
Africa 3
… …
Asia 8
Europe 5
Africa 3
… …
Asia 7
Europe 6
KTable usersPerRegion KTable usersPerRegion

46Confidential
4
7
5
3
2
8 4
7
6
3
2
7
Alice
Real-time dashboard
alice Europe
user-locations
… …
… …
-1
+1
user-locations
(mobile team)
user-prefs
(web team)

47Confidential
Streams meet Tables – in the DSL

48Confidential
Streams meet Tables
• Most use cases for stream processing require both Streams and Tables
• Essential for any stateful computations
• Kafka ships with first-class support for Streams and Tables
• Scalability, fault tolerance, efficient joins and aggregations, …
• Benefits include: simplified architectures, less moving pieces, less Do-It-Yourself work

50Confidential
Key features in 0.10
• Native, 100%-compatible Kafka integration

51Confidential
Native, 100% compatible Kafka integration
Read from Kafka
Write to Kafka

52Confidential
• Secure stream processing using Kafka’s security features

53Confidential
Secure stream processing with the Streams API
• Your applications can leverage all client-side security features in Apache Kafka
• Security features include:
• Encrypting data-in-transit between applications and Kafka clusters
• Authenticating applications against Kafka clusters (“only some apps may talk to the production
cluster”)
• Authorizing application against Kafka clusters (“only some apps may read data from sensitive topics”)

54Confidential
Configuring security settings
• In general, you can configure both Kafka Streams plus the underlying Kafka clients in your
apps

55Confidential
Configuring security settings
• Example: encrypting data-in-transit + client authentication to Kafka cluster
Full demo application at https://github.com/confluentinc/examples

56Confidential
• Elastic and highly scalable
• Fault-tolerant

60Confidential
• Fault-tolerant
• Stateful and stateless computations

61Confidential
Stateful computations
• Stateful computations like aggregations (e.g. counting), joins, or windowing require state
• State stores are the backbone of state management
• … are local for best performance
• … are backed up to Kafka for elasticity and for fault-tolerance
• ... are per stream task for isolation – think: share-nothing
• Pluggable storage engines
• Default: RocksDB (a key-value store) to allow for local state that is larger than available RAM
• You can also use your own, custom storage engine
• From the user perspective:
• DSL: no need to worry about anything, state management is automatically being done for you
• Processor API: direct access to state stores – very flexible but more manual work

66Confidential
Use case: real-time, distributed joins at large scale

67Confidential

68Confidential

69Confidential
Stateful computations
• Use the Processor API to interact directly with state stores
Get the store
Use the store

70Confidential
• Fault-tolerant
• Interactive queries

72Confidential
Interactive Queries: architecture comparison
Kafka
Streams
App
App
App
App
1 Capture business
events in Kafka
2 Process the events
with Kafka Streams
4
Other apps query external
systems for latest results
! Must use external systems
to share latest results
App
App
App
1 Capture business
events in Kafka
2 Process the events
with Kafka Streams
3 Now other apps can directly
query the latest results
Before (0.10.0)
After (0.10.1): simplified, more app-centric architecture
Kafka
Streams
App

73Confidential
• Fault-tolerant
• Time model

76Confidential
Time
• You configure the desired time semantics through timestamp extractors
• Default extractor yields event-time semantics
• Extracts embedded timestamps of Kafka messages (introduced in v0.10)

77Confidential
• Fault-tolerant
• Time model
• Windowing

78Confidential
Windowing
• Group events in a stream using time-based windows
• Use case examples:
• Time-based analysis of ad impressions (”number of ads clicked in the past hour”)
• Monitoring statistics of telemetry data (“1min/5min/15min averages”)
Input data, where
colors represent
different users events
Rectangles denote
different event-time
windows
processing-time
event-time
windowing
alice
bob
dave

79Confidential
Windowing in the DSL
TimeWindows.of(3000)
TimeWindows.of(3000).advanceBy(1000)

80Confidential
• Fault-tolerant
• Time model
• Windowing
• Supports late-arriving and out-of-order data

81Confidential
Out-of-order and late-arriving data
• Is very common in practice, not a rare corner case
• Related to time model discussion

82Confidential
Out-of-order and late-arriving data: example when this will happen
Users with mobile phones enter
airplane, lose Internet connectivity
Emails are being written
during the 10h flight
Internet connectivity is restored,
phones will send queued emails now

83Confidential
Out-of-order and late-arriving data
• Is very common in practice, not a rare corner case
• Related to time model discussion
• We want control over how out-of-order data is handled, and handling must be efficient
• Example: We process data in 5-minute windows, e.g. compute statistics
• Option A: When event arrives 1 minute late: update the original result!
• Option B: When event arrives 2 hours late: discard it!

84Confidential
• Fault-tolerant
• Time model
• Windowing
• Supports late-arriving and out-of-order data
• Millisecond processing latency, no micro-batching
• At-least-once processing guarantees (exactly-once is in the works as we speak)

85Confidential
Roadmap Outlook

86Confidential
Roadmap outlook for Kafka Streams
• Exactly-Once processing semantics
• Unified API for real-time processing and “batch” processing
• Global KTables
• Session windows
• … and more …

88Confidential
Where to go from here
• Kafka Streams is available in Confluent Platform 3.1 and in Apache Kafka 0.10.1
• http://www.confluent.io/download
• Kafka Streams demos: https://github.com/confluentinc/examples
• Java 7, Java 8+ with lambdas, and Scala
• WordCount, Interactive Queries, Joins, Security, Windowing, Avro integration, …
• Confluent documentation: http://docs.confluent.io/current/streams/
• Quickstart, Concepts, Architecture, Developer Guide, FAQ
• Recorded talks
• Introduction to Kafka Streams:
http://www.youtube.com/watch?v=o7zSLNiTZbA
• Application Development and Data in the Emerging World of Stream Processing (higher level talk):
https://www.youtube.com/watch?v=JQnNHO5506w

90Confidential
Appendix: Streams and Tables
A closer look

91Confidential
Motivating example: continuously compute current users per geo-region
4
7
5
3
2
8
Real-time dashboard
… …
user-locations
(mobile team)
user-prefs
(web team)

92Confidential
4
7
5
3
2
8
Real-time dashboard
alice Europe
user-locations
… …
user-locations
(mobile team)
user-prefs
(web team)

93Confidential
4
7
5
3
2
8
Real-time dashboard
alice Europe
user-locations
user-locations
(mobile team)
user-prefs
(web team)
… …
… …

94Confidential
4
7
5
3
2
8 4
7
6
3
2
7
Alice
Real-time dashboard
alice Europe
user-locations
… …
… …
-1
+1
user-locations
(mobile team)
user-prefs
(web team)

95Confidential
Same data, but different use cases require different interpretations
alice San Francisco
alice New York City
alice Rio de Janeiro
alice Sydney
alice Beijing
alice Paris
alice Berlin

96Confidential
alice San Francisco
alice New York City
alice Sydney
alice Beijing
alice Paris
alice Berlin
Use case 1: Frequent traveler status?
Use case 2: Current location?

97Confidential
“Alice has been to SFO, NYC, Rio, Sydney,
Beijing, Paris, and finally Berlin.”
“Alice is in SFO, NYC, Rio, Sydney,
Beijing, Paris, Berlin right now.”
⚑ ⚑
⚑⚑
⚑
⚑
⚑ ⚑ ⚑
⚑⚑
⚑
⚑
⚑
Use case 1: Frequent traveler status? Use case 2: Current location?

98Confidential
alice San Francisco
alice New York City
alice Sydney
alice Beijing
alice Paris
alice Berlin
⚑ ⚑ ⚑⚑
⚑
⚑⚑
⚑

99Confidential
alice San Francisco
alice New York City
alice Sydney
alice Beijing
alice Paris
alice Berlin
⚑ ⚑ ⚑⚑
⚑
⚑⚑
⚑

100Confidential
alice San Francisco
alice New York City
alice Sydney
alice Beijing
alice Paris
alice Berlin
⚑ ⚑ ⚑⚑
⚑
⚑⚑
⚑

101Confidential
Streams meet Tables
record stream
When you need… so that the topic is
interpreted as a
All the values of a key KStream
then you’d read the
Kafka topic into a
Example
All the places Alice
has ever been to
with messages
interpreted as
INSERT
(append)

102Confidential
Streams meet Tables
record stream
changelog stream
When you need… so that the topic is
interpreted as a
All the values of a key
Latest value of a key
KStream
KTable
then you’d read the
Kafka topic into a
Example
All the places Alice
has ever been to
Where Alice
is right now
with messages
interpreted as
INSERT
(append)
UPSERT
(overwrite
existing)

103Confidential
“Alice has been to SFO, NYC, Rio, Sydney,
Beijing, Paris, and finally Berlin.”
“Alice is in SFO, NYC, Rio, Sydney,
Beijing, Paris, Berlin right now.”
⚑ ⚑
⚑⚑
⚑
⚑
⚑ ⚑ ⚑
⚑⚑
⚑
⚑
⚑
Use case 1: Frequent traveler status? Use case 2: Current location?
KStream KTable

104Confidential
4
7
5
3
2
8 4
7
6
3
2
7
Alice
Real-time dashboard
alice Europe
user-locations
… …
… …
-1
+1
user-locations
(mobile team)
user-prefs
(web team)

105Confidential

106Confidential
alice Europe
user-locations
… …
… …
KTable userProfilesKTable userProfiles

107Confidential
// Compute per-‐region statistics (continuously updated)
KTable<UserId, Long> usersPerRegion = userProfiles
.filter((userId, profile) -‐> profile.age < 30)
.groupBy((userId, profile) -‐> profile.location)
.count();
alice Europe
user-locations
Africa 3
… …
Asia 8
Europe 5
Africa 3
… …
Asia 7
Europe 6
KTable usersPerRegion KTable usersPerRegion

108Confidential
4
7
5
3
2
8 4
7
6
3
2
7
Alice
Real-time dashboard
alice Europe
user-locations
… …
… …
-1
+1
user-locations
(mobile team)
user-prefs
(web team)

109Confidential
Another common use case: continuous transformations
• Example: to enrich an input stream (user clicks) with side data (current user profile)
KStream alice /rental/p8454vb, 06:59 PM PDT
user-clicks-topics (at 1M msgs/s)
“facts”

110Confidential
KStream alice /rental/p8454vb, 06:59 PM PDT
alice Asia, 25y
bob Europe, 46y
… …
KTable
user-profiles-topic
“facts”
“dimensions”

111Confidential
KStream
alice /rental/p8454vb, 06:59 PDT, Asia, 25y
stream.JOIN(table)
alice /rental/p8454vb, 06:59 PM PDT
alice Asia, 25y
bob Europe, 46y
… …
KTable
user-profiles-topic
“facts”
“dimensions”

112Confidential
KStream
alice /rental/p8454vb, 06:59 PDT, Asia, 25y
stream.JOIN(table)
alice /rental/p8454vb, 06:59 PM PDT
alice Asia, 25y
bob Europe, 46y
… …
KTable
alice Europe, 25y
bob Europe, 46y
… …alice Europe
new update for alice from user-locations topic
user-profiles-topic
“facts”
“dimensions”

113Confidential
Appendix: Interactive Queries
A closer look

114Confidential
Interactive Queries

115Confidential
Interactive Queries
charlie 3bob 5 alice 2

116Confidential
Interactive Queries
New API to access
local state stores of
an app instance

117Confidential
Interactive Queries
New API to discover
running app instances
“host1:4460” “host5:5307” “host3:4777”

118Confidential
Interactive Queries
You: inter-app communication (RPC layer)

Introducing Kafka's Streams API

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Viewers also liked

Viewers also liked (20)

Similar to Introducing Kafka's Streams API

Similar to Introducing Kafka's Streams API (20)

More from confluent

More from confluent (20)

Recently uploaded

Recently uploaded (20)

Introducing Kafka's Streams API