1.
Till Rohrmann
till@data-artisans.com
@stsffap
From Apache Flink®
1.3 to 1.4

2.
2
Original creators of Apache
Flink®
Providers of
dA Platform 2, including
open source Apache Flink +
dA Application Manager

3.
Overview
Apache Flink 1.3 – Previously on Apache
Flink
Apache Flink 1.4 – What’s happening now?
Apache Flink 1.5+ – Next on Apache Flink
3

4.
Previously on Apache Flink
Apache Flink 1.3

5.
Apache Flink 1.3 in Numbers
141 contributors (no deduplication)
1400 commits
>= 680 resolved JIRA issues
+261813 / -65646 LOC
5

6.
Evolution of Flink’s API
6
Flink 1.0.0
State API (ValueState
ReducingState, ListState)
Flink 1.1.0
Session Windows
Late arriving events
Flink 1.2.0
ProcessFunction (access
to state, timers, events)
Flink 1.3.0
Side outputs
Access to per-window state

7.
Side Outputs
 Additional outputs for a stream
 Late events
 Corrupted input data
 More expressive APIs
 FLINK-4460
7
Process
Function
Main output
Side output

8.
Side Outputs: Example
8
DataStream<Integer> input = ...;
final OutputTag<String> outputTag = new OutputTag<String>("side-output"){};
SingleOutputStreamOperator<Integer> mainDataStream = input
.process(new ProcessFunction<Integer, Integer>() {
@Override
public void processElement(
Integer value,
Context ctx,
Collector<Integer> out) throws Exception {
// emit data to regular output
out.collect(value);
// emit data to side output
ctx.output(outputTag, "sideout-" + String.valueOf(value));
}
});
DataStream<String> sideOutputStream = mainDataStream.getSideOutput(outputTag);

9.
Evolution of Large State Handling
9
Flink 1.0.0
RocksDB for out-of-core
state support
Flink 1.1.0
Fully async RocksDB
snapshots
Flink 1.2.0
Rescalable keyed and
non-partitioned state
Flink 1.3.0
Incremental checkpoints
Fine-grained recovery

10.
G
H
C
D
Full Checkpoints
10
Checkpoint 1 Checkpoint 2 Checkpoint 3
I
E
A
B
C
D
A
B
C
D
A
F
C
D
E
@t1 @t2 @t3
A
F
C
D
E
G
H
C
D
I
E

11.
G
H
C
D
Incremental Checkpoints
11
Checkpoint 1 Checkpoint 2 Checkpoint 3
I
E
A
B
C
D
A
B
C
D
A
F
C
D
E
E
F
G
H
I
@t1 @t2 @t3

12.
Incremental Checkpoints
12
Checkpoint 1 Checkpoint 2 Checkpoint 3 Checkpoint 4
C1 C3C1 C1
Chunk
1
Chunk
2
Chunk
3
Chunk
4
Storage
C2 C4C3

13.
Incremental Checkpointing Contd.
Currently supported for RocksDB
state backend
FLINK-5053
Faster and smaller checkpoints
13
Full checkpoint Incremental checkpoint
Size 60 GB 1 – 30 GB
Time 180 s 3 – 30 s
“A Look at Flink’s Internal
Data Structures and
Algorithms for Efficient
Checkpointing” by Stefan
Richter, Tomorrow @
12:20 pm Maschinenhaus

14.
Evolution of High Level APIs
14
Flink 1.0.0
CEP library added
Table API v1
Flink 1.1.0
Table API overhaul
Integration with Apache Calcite
Flink 1.2.0
Tumbling, sliding and session
group-windows for Table API
Flink 1.3.0
Rescalable CEP operators
Retractions in Table API/SQL

15.
Enriched CEP Language
Support for quantifiers (+, *, ?)
FLINK-3318
Iterative conditions
FLINK-6197
Not operator
FLINK-3320
15
“Complex Event Processing With
Flink: The State of FlinkCEP” by
Kostas Kloudas, Today @ 2:30
pm Maschinenhaus

16.
CEP: Detect Dipping Stocks
16
DataStream<Stock> stocks = …;
Pattern<Stock, ?> pattern = Pattern
.<Stock>begin("rising")
.where(new IterativeCondition<Stock>() {
@Override
public boolean filter(Stock stock, Context<Stock> ctx) throws Exception {
// calculate the average price
double sum = 0.0; int count = 0;
for (Stock previousStock : ctx.getEventsForPattern("rising")) {
sum += previousStock.getPrice(); count++;
}
// only accept if the price is higher or equal than the average price
return stock.getPrice() >= sum / count;
}).oneOrMore()
.next("falling");
PatternStream<Stock> dippingStocks = new PatternStream<>(stocks.keyBy("name"), pattern);
DataStream<String> namesOfDippingStocks = dippingStocks.select(…);

17.
What’s Happening Now?
Apache Flink 1.4

18.
Event Driven I/O
18
Rework of Flink’s network stack
Event driven network I/O
Use full available capacity
Near perfect latency behaviour
TCP
Buffer
capacity left
flush

19.
Flow Control
 Flow control for TaskManager communication
 Single channel no longer stalls other
multiplexed channels
 Fine-grained backpressure control
 Improves checkpoint alignments
19
“Building a Network Stack
for Optimal Throughput /
Low-Latency Trade-Offs”
by Nico Kruber, Today @
2:00 pm Palais Atelier
Receiver
Sender #1
Sender #2
Give credit
Send
credited data

20.
New Deployment Model
Rework of Flink’s distributed
architecture
Ready for multitude of
deployment scenarios
Support for dynamic scaling
20
“Flink in Containerland” by
Patrick Lucas, Tomorrow
@ 3:20 pm Maschinenhaus

21.
Producing Exactly Once with Kafka 0.11
Support for Kafka 0.11
First Kafka producer with
exactly once processing
guarantees
21
“Hit Me, Baby, Just One Time
– Building End-to-End Exactly
Once Applications With Flink”
by Piotr Nowojski, Today @
3:20 pm Palais Atelier
Consuming Producing
End-to-End exactly once processing

22.
StreamSQL and Table API
Support for retractions
Extended aggregation support
Support for external table
catalogs
Window joins
22
“Unified Stream and Batch
Processing With Apache
Flink’s Relational APIs” by
Fabian Hüske, Tomorrow
@ 11:00 am Kesselhaus
“From Streams to Tables
and Back Again: A Demo
of Flink’s Table & SQL
API” by Timo Walther,
Tomorrow @ 11:50 am
Kesselhaus

23.
Operational Robustness
Drop Java 7
Support Scala 2.12
Avoid dependency hell
Child first class loading
Relocation of
dependencies
De-Hadoopification
23

24.
Next on Apache Flink
Apache Flink 1.5+

25.
Side Inputs
 Additional input for operator
 Join with static data set
 Feeding of externally trained ML model
 Window joins
 Flip-17 design document: https://goo.gl/W4yMEu
25
Process
Function
Main input
Side input

26.
State Management & Evolution
Eager state declaration
State type, serializer and name
known at pre-flight time
Flip-22 design document:
https://goo.gl/trFiSi
Evolving existing state
Schema updates
Serializer upgrades
26
“Managing State in
Apache Flink” by
Tzu-Li Tai, Today @
4:30 pm Kesselhaus

27.
State Replication
Replicate state between
TaskManagers
Faster recovery in
case of failures
High throughput
queryable state
27
TaskManager
TaskManager
Change log stream
Input
State

28.
Programmatic Job Control
Improve client to give better job control
Run concurrent jobs from the same
program
Trigger savepoints programmatically
Better testing facilities
28

29.
JobClient & ClusterClient
29
StreamExecutionEnvironment env = ...;
// define program
JobClient jobClient = env.execute();
CompletableFuture<Acknowledge> savepointFuture = jobClient.takeSavepoint(savepointPath);
// wait for the savepoint completion
savepointFuture.get();
CompletableFuture<JobExecutionResult> resultFuture = jobClient.getResultFuture();
// cancel the job
jobClient.cancelJob();
// get the execution result --> should be canceled
JobExecutionResult result = resultFuture.get();
// get list of all still running jobs on the cluster
ClusterClient clusterClient = jobClient.getClusterClient();
CompletableFuture<List<JobInfo>> jobInfosFuture = clusterClient.getJobInfos();
List<JobInfo> jobInfos = jobInfosFuture.get();

30.
TL;DL
Apache Flink one of the most innovative open
source stream processing platforms
Stay tuned what’s happening next 
Visit the in depths talks to learn more about
Flink’s internals
30

31.
31
Thank you!
@stsffap
@ApacheFlink
@dataArtisans

32.
We are hiring!
data-artisans.com/careers
32