Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Kafka Partitioning Algorithm
Search
Gordon Diggs
September 02, 2016
Technology
0
130
Kafka Partitioning Algorithm
A look at the default partitioning algorithm that Kafka uses
Gordon Diggs
September 02, 2016
Tweet
Share
More Decks by Gordon Diggs
See All by Gordon Diggs
John Coltrane: Lessons in Leadership
gordondiggs
2
280
The Customer Gap
gordondiggs
1
100
Picking Records with JavaScript and a Button
gordondiggs
0
74
Supbutton
gordondiggs
0
59
Rayons
gordondiggs
0
61
Sous Vide
gordondiggs
0
77
Dev Events & Internal Tools at Paperless Post
gordondiggs
0
120
The Joys and Pains of Working With an Old Codebase
gordondiggs
0
120
The Joys and Pains of Working with an Old Codebase
gordondiggs
1
2.3k
Other Decks in Technology
See All in Technology
C++26アップデート 2025-03
faithandbrave
0
1.1k
Running JavaScript within Ruby
hmsk
3
400
【Λ(らむだ)】最近のアプデ情報 / RPALT20250422
lambda
0
130
ドキュメント管理の理想と現実
kazuhe
1
250
AIと共に乗り越える、 入社後2ヶ月の苦労と学習の軌跡
sai_kaneko
0
130
30代からでも遅くない! 内製開発の世界に飛び込み、最前線で戦うLLMアプリ開発エンジニアになろう
minorun365
PRO
15
4.7k
От ручной разметки к LLM: как мы создавали облако тегов в Lamoda. Анастасия Ангелова, Data Scientist, Lamoda Tech
lamodatech
0
820
2025-04-24 "Manga AI Understanding & Localization" Furukawa Arata (CyberAgent, Inc)
ornew
2
300
SREからゼロイチプロダクト開発へ ー越境する打席の立ち方と期待への応え方ー / Product Engineering Night #8
itkq
2
1.1k
Perl歴約10年のエンジニアがフルスタックTypeScriptに出会ってみた
papix
1
200
PicoRabbit: a Tiny Presentation Device Powered by Ruby
harukasan
PRO
2
270
OpsJAWS34_CloudTrailLake_for_Organizations
hiashisan
0
160
Featured
See All Featured
Git: the NoSQL Database
bkeepers
PRO
430
65k
What’s in a name? Adding method to the madness
productmarketing
PRO
22
3.4k
The Psychology of Web Performance [Beyond Tellerrand 2023]
tammyeverts
47
2.7k
I Don’t Have Time: Getting Over the Fear to Launch Your Podcast
jcasabona
32
2.2k
We Have a Design System, Now What?
morganepeng
52
7.5k
How to train your dragon (web standard)
notwaldorf
90
6k
Bash Introduction
62gerente
611
210k
Why You Should Never Use an ORM
jnunemaker
PRO
56
9.3k
Measuring & Analyzing Core Web Vitals
bluesmoon
7
400
Designing Dashboards & Data Visualisations in Web Apps
destraynor
231
53k
Six Lessons from altMBA
skipperchong
28
3.7k
Building Adaptive Systems
keathley
41
2.5k
Transcript
Kafka Partitioning Algorithm CC LnL 20160902
Kafka Partitioning Algorithm CC LnL 20160902
Kafka Partitioning • Can be keyed • Round robin (mostly)
by default
public int partition(String topic, Object key, byte[] keyBytes, Object value,
byte[] valueBytes, Cluster cluster) { List<PartitionInfo> partitions = cluster.partitionsForTopic(topic); int numPartitions = partitions.size(); if (keyBytes == null) { int nextValue = counter.getAndIncrement(); List<PartitionInfo> availablePartitions = cluster.availablePartitionsForTopic(topic); if (availablePartitions.size() > 0) { int part = Utils.toPositive(nextValue) % availablePartitions.size(); return availablePartitions.get(part).partition(); } else { // no partitions are available, give a non-available partition return Utils.toPositive(nextValue) % numPartitions; } } else { // hash the keyBytes to choose a partition return Utils.toPositive(Utils.murmur2(keyBytes)) % numPartitions; } }
public int partition(String topic, Object key, byte[] keyBytes, Object value,
byte[] valueBytes, Cluster cluster) { List<PartitionInfo> partitions = cluster.partitionsForTopic(topic); int numPartitions = partitions.size(); if (keyBytes == null) { int nextValue = counter.getAndIncrement(); List<PartitionInfo> availablePartitions = cluster.availablePartitionsForTopic(topic); if (availablePartitions.size() > 0) { int part = Utils.toPositive(nextValue) % availablePartitions.size(); return availablePartitions.get(part).partition(); } else { // no partitions are available, give a non-available partition return Utils.toPositive(nextValue) % numPartitions; } } else { // hash the keyBytes to choose a partition return Utils.toPositive(Utils.murmur2(keyBytes)) % numPartitions; } }
public int partition(String topic, Object key, byte[] keyBytes, Object value,
byte[] valueBytes, Cluster cluster) { List<PartitionInfo> partitions = cluster.partitionsForTopic(topic); int numPartitions = partitions.size(); if (keyBytes == null) { int nextValue = counter.getAndIncrement(); List<PartitionInfo> availablePartitions = cluster.availablePartitionsForTopic(topic); if (availablePartitions.size() > 0) { int part = Utils.toPositive(nextValue) % availablePartitions.size(); return availablePartitions.get(part).partition(); } else { // no partitions are available, give a non-available partition return Utils.toPositive(nextValue) % numPartitions; } } else { // hash the keyBytes to choose a partition return Utils.toPositive(Utils.murmur2(keyBytes)) % numPartitions; } }
public int partition(String topic, Object key, byte[] keyBytes, Object value,
byte[] valueBytes, Cluster cluster) { List<PartitionInfo> partitions = cluster.partitionsForTopic(topic); int numPartitions = partitions.size(); if (keyBytes == null) { int nextValue = counter.getAndIncrement(); List<PartitionInfo> availablePartitions = cluster.availablePartitionsForTopic(topic); if (availablePartitions.size() > 0) { int part = Utils.toPositive(nextValue) % availablePartitions.size(); return availablePartitions.get(part).partition(); } else { // no partitions are available, give a non-available partition return Utils.toPositive(nextValue) % numPartitions; } } else { // hash the keyBytes to choose a partition return Utils.toPositive(Utils.murmur2(keyBytes)) % numPartitions; } }
public int partition(String topic, Object key, byte[] keyBytes, Object value,
byte[] valueBytes, Cluster cluster) { List<PartitionInfo> partitions = cluster.partitionsForTopic(topic); int numPartitions = partitions.size(); if (keyBytes == null) { int nextValue = counter.getAndIncrement(); List<PartitionInfo> availablePartitions = cluster.availablePartitionsForTopic(topic); if (availablePartitions.size() > 0) { int part = Utils.toPositive(nextValue) % availablePartitions.size(); return availablePartitions.get(part).partition(); } else { // no partitions are available, give a non-available partition return Utils.toPositive(nextValue) % numPartitions; } } else { // hash the keyBytes to choose a partition return Utils.toPositive(Utils.murmur2(keyBytes)) % numPartitions; } }
public int partition(String topic, Object key, byte[] keyBytes, Object value,
byte[] valueBytes, Cluster cluster) { List<PartitionInfo> partitions = cluster.partitionsForTopic(topic); int numPartitions = partitions.size(); if (keyBytes == null) { int nextValue = counter.getAndIncrement(); List<PartitionInfo> availablePartitions = cluster.availablePartitionsForTopic(topic); if (availablePartitions.size() > 0) { int part = Utils.toPositive(nextValue) % availablePartitions.size(); return availablePartitions.get(part).partition(); } else { // no partitions are available, give a non-available partition return Utils.toPositive(nextValue) % numPartitions; } } else { // hash the keyBytes to choose a partition return Utils.toPositive(Utils.murmur2(keyBytes)) % numPartitions; } }
public int partition(String topic, Object key, byte[] keyBytes, Object value,
byte[] valueBytes, Cluster cluster) { List<PartitionInfo> partitions = cluster.partitionsForTopic(topic); int numPartitions = partitions.size(); if (keyBytes == null) { int nextValue = counter.getAndIncrement(); List<PartitionInfo> availablePartitions = cluster.availablePartitionsForTopic(topic); if (availablePartitions.size() > 0) { int part = Utils.toPositive(nextValue) % availablePartitions.size(); return availablePartitions.get(part).partition(); } else { // no partitions are available, give a non-available partition return Utils.toPositive(nextValue) % numPartitions; } } else { // hash the keyBytes to choose a partition return Utils.toPositive(Utils.murmur2(keyBytes)) % numPartitions; } }
public int partition(String topic, Object key, byte[] keyBytes, Object value,
byte[] valueBytes, Cluster cluster) { List<PartitionInfo> partitions = cluster.partitionsForTopic(topic); int numPartitions = partitions.size(); if (keyBytes == null) { int nextValue = counter.getAndIncrement(); List<PartitionInfo> availablePartitions = cluster.availablePartitionsForTopic(topic); if (availablePartitions.size() > 0) { int part = Utils.toPositive(nextValue) % availablePartitions.size(); return availablePartitions.get(part).partition(); } else { // no partitions are available, give a non-available partition return Utils.toPositive(nextValue) % numPartitions; } } else { // hash the keyBytes to choose a partition return Utils.toPositive(Utils.murmur2(keyBytes)) % numPartitions; } }
Utils.toPositive
public static int toPositive(int number) { return number & 0x7fffffff;
}
0111 1111 1111 1111 1111 1111 1111 1111
0000 0000 0000 0000 0000 0000 0000 1010 & 0111
1111 1111 1111 1111 1111 1111 1111 = 0000 0000 0000 0000 0000 0000 0000 1010
Two’s Complement
0000 0000 0000 0000 0000 0000 0000 1010 1111 1111
1111 1111 1111 1111 1111 0101 1111 1111 1111 1111 1111 1111 1111 1010
1111 1111 1111 1111 1111 1111 1111 1010 & 0111
1111 1111 1111 1111 1111 1111 1111 = 0111 1111 1111 1111 1111 1111 1111 1010
None