Cassandra introduction 2016

Introduction to Cassandra
DuyHai DOAN
Apache Cassandra Evangelist

Datastax
•  Founded in April 2010
•  We contribute a lot to Apache Cassandra™
•  400+ customers (25 of the Fortune 100), 450+ employees
•  Headquarter in San Francisco Bay area
•  EU headquarter in London, ofﬁces in France and Germany
•  Datastax Enterprise = OSS Cassandra + extra features
© 2016 DataStax, All Rights Reserved.
2

Cassandra history
•  created at Facebook
•  open-sourced since 2008
•  current version: 3.2
•  column-oriented ☞ distributed table
3

5 Cassandra key points
•  Linear scalability
•  Continuous availability
•  Multi Data-center native
•  Operational simplicity
•  Spark integration
4

1) Linear scalability
5
C*
C* C*
NetcoSports
3 nodes, ≈3GB
1k+ nodes, PB+
YOU

2) Continuous availability
6
•  thanks to the Dynamo architecture

3) Multi Data-centers
7
•  out-of-the-box (conﬁg only)
•  AWS conﬁg for multi-regions DCs
•  GCE support
•  Microsoft Azure support
•  CloudStack support

Multi DC usages
Data locality, disaster recovery
8
C*
C*
C*
C*
C* C*
C* C* C*
C*
C*
C*
C*
New York (DC1) London (DC2)
Async
replication

Multi DC usages
Virtual DC for workload segregation
9
C*
C*
C*
C*
C* C*
C* C* C*
C*
C*
C*
C*
Production
(LIVE)
Analytics
(Spark)
Async
replication
Same room

Multi DC usages
Prod data copy for back-up/benchmark
10
C*
C*
C*
C*
C* C*
C* C* C*
C*
C*
C*
C*
Use
LOCAL_XXX
Consistency
Levels
My tiny test DC
READ-ONLY!!!
Async
replication

4) Operational simplicity
11
•  1 node = 1 process + 2 conﬁg ﬁles (cassandra.yaml + cassandra-rackdc.properties)
•  deployment automation
•  OpsCenter for
•  monitoring
•  provisioning*
•  services* (repair, performance, …)
* only with Datastax Enterprise

4) Operational simplicity
12

5) Spark integration
13
•  Cassandra + Spark = awesome !
•  Spark/Cassandra connector = most advanced connector right now for NoSQL
db
•  predicates push-down
•  early ﬁltering
•  dataframe integration
•  Analytics, aggregation, streaming …

Main Cassandra use-cases
14

Cassandra use-cases
15
Messaging
Collections/
Playlists
Fraud
detection
Recommendation/
Personalization
Internet of things/
Sensor data

Cassandra use-cases
16
Messaging
Collections/
Playlists
Fraud
detection
Recommendation/
Personalization
Internet of things/
Sensor data

17
Q & A
! "

Layers
18
•  Cluster
•  Amazon DynamoDB paper
•  masterless
•  Storage engine
•  Google Big Table
•  columns/columns family ☞ distributed tables

Data Distribution
19

The tokens
20
Random hash of #partition à token = hash(#p)
Hash: ] –x, x ]
hash range: 264 values
x = 264/2
C*
C*
C*
C*
C* C*
C* C*

Token ranges
21
A: −x,−
3x
4
⎤
⎦
⎥
⎥
⎤
⎦
⎥
⎥
B: −
3x
4
,−
2x
4
⎤
⎦
⎥
⎥
⎤
⎦
⎥
⎥
C: −
2x
4
,−
x
4
⎤
⎦
⎥
⎥
⎤
⎦
⎥
⎥
D: −
x
4
,0
⎤
⎦
⎥
⎥
⎤
⎦
⎥
⎥
E: 0,
x
4
⎤
⎦
⎥
⎥
⎤
⎦
⎥
⎥
F:
x
4
,
2x
4
⎤
⎦
⎥
⎥
⎤
⎦
⎥
⎥
G:
2x
4
,
3x
4
⎤
⎦
⎥
⎥
⎤
⎦
⎥
⎥
H :
3x
4
,x
⎤
⎦
⎥
⎥
⎤
⎦
⎥
⎥
C*
C*
C*
C*
C* C*
C* C*

Distributed tables
22
H
A
E
D
B C
G F
user_id1
user_id2
user_id3
user_id4
user_id5
CREATE TABLE users(
user_id int,
…,
PRIMARY KEY(user_id)
),

Distributed tables
23
H
A
E
D
B C
G F
user_id1
user_id2
user_id3
user_id4
user_id5

Linear scalability
24
H
A
E
D
B C
G F
Today = high load
•  disk occupation 80%
•  CPU 70%
•  saturated memory

Scaling out
25
H
A
E
D
B
C
G
F
I
J
+2 nodes
•  disk occupation 50%
•  CPU 50%
•  memory ✌︎
Automatic data rebalancing
•  each node gives up some tokens
•  flag to throttle network bandwidth
•  streamingthroughput

Automatic data re-balancing with virtual nodes
26
A:
B:
C:
D:
E:
F:
G:
H:
A:
B:
C:
D:
E:
F:
G:
H:
I:
J:
+2 nodes

27
Q & A
! "

Replication Model & Consistency
28

Failure tolerance
29
Replication factor (RF) = 3
H
A
E
D
B C
G F
1
2 3
{A, H, G}
{B, A, H} {C, B, A}

Coordinator node
30
Responsible for handling requests (read/write)
Every node can be coordinator
•  masterless
•  round robin master for each request
•  no SPOF
•  proxy role
H
A
E
D
B C
G F
coordinator
request
1
2 3

Consistency level
31
Tunable at runtime
•  ONE
•  QUORUM (strict majority w.r.t RF)
•  ALL
Applicable to any request (read/write)

Consistency in action
32
B A A
B A A
Read ONE: A
data replication in progress …
Write ONE: B
ack
RF = 3, Write ONE, Read ONE

33
B A A
B A A
Read QUORUM: A
Write ONE: B
ack
RF = 3, Write ONE, Read QUORUM

34
B A A
B A A
Read ALL: B
Write ONE: B
ack
RF = 3, Write ONE, Read ALL

35
B B A
B B A
Read ONE: A
Write QUORUM: B
ack
RF = 3, Write QUORUM, Read ONE

36
B B A
B B A
Read QUORUM: A
Write QUORUM: B
ack
RF = 3, Write QUORUM, Read QUORUM

Consistency level = trade-oﬀ
37

Consistency level
38
ONE
Fast, may not read latest written value

Consistency level
39
QUORUM
Strict majority w.r.t. Replication Factor
Good balance

Consistency level
40
ALL
Paranoid
Slow, lost of high availability

Consistency level common patterns
41
ONERead + ONEWrite
☞ available for read/write even (N-1) replicas down
QUORUMRead + QUORUMWrite
☞ available for read/write even if (RF - 1) replica (s) down

42
Q & A
! "

Last Write Win & Compaction
43

Last Write Win (LWW)
44
jdoe
age name
33 John DOE
INSERT INTO users(login, name, age) VALUES('jdoe', 'John DOE', 33);
#partition

45
jdoe
age (t1) name (t1)
33 John DOE
auto-generated timestamp (μs)
.

46
UPDATE users SET age = 34 WHERE login = 'jdoe';
jdoe
age (t1) name (t1)
33 John DOE
jdoe
age (t2)
34
SSTable1 SSTable2

47
DELETE age FROM users WHERE login = 'jdoe';
jdoe
age (t1) name (t1)
33 John DOE
jdoe
age (t2)
34
SSTable1 SSTable2
tombstone
SSTable3
jdoe
age (t3)
ý

48
SELECT age FROM users WHERE login = 'jdoe';
jdoe
age (t1) name (t1)
33 John DOE
jdoe
age (t2)
34
SSTable1 SSTable2 SSTable3
jdoe
age (t3)
ý
???

49
jdoe
age (t1) name (t1)
33 John DOE
jdoe
age (t2)
34
jdoe
age (t3)
ý
✓✕✕

Compaction
50
jdoe
age (t3)
ý
jdoe
age (t1) name (t1)
33 John DOE
jdoe
age (t2)
34
New SSTable
jdoe
age (t3) name (t1)
ý John DOE

Basic Data Modeling
51

Table creation
52
CREATE TABLE users (
login text,
name text,
age int,
…
PRIMARY KEY(login));
partition key (#partition)

DML statements
53
UPDATE users SET age = 34 WHERE login = 'jdoe';
DELETE age FROM users WHERE login = 'jdoe';

What’s about joins ?
54
How can I join data between tables ?
How can I model 1 – N relationships ?
How to model a mailbox ?
EmailsUser
1 n

Compound primary key
55
CREATE TABLE mailbox (
login text,
message_id timeuuid,
interlocutor text,
message text,
PRIMARY KEY((login), message_id));
partition key clustering column unicity

Compound primary key
56
rsmith
2014-11-21 16:00:00
‘bobm’, ‘It’s really…’
2014-11-21 17:32:12
‘bobm’, ‘It depends..’
2014-11-21 21:21:09
‘bobm’, ‘Don’t do…’
…
hsue
2014-11-21 11:04:43
‘jdoe’, ‘Hi, …’
2014-11-21 11:22:43
‘rsmith’, ‘Hello,…’
jdoe
2014-11-21 11:00:00
‘hsue’, ‘Hi there!’
2014-11-21 11:22:43
‘rsmith’, ‘Hello,…’
2014-11-21 13:06:19
‘bobm’, ‘Do you…’
ordered by clustering column (date)
Not
ordered

Queries
57
Get message by user and message_id (date)
Get message by user and date interval
SELECT * FROM mailbox WHERE login = 'jdoe'
and message_id = ‘2014-11-21 16:00:00’;
SELECT * FROM mailbox WHERE login = 'jdoe'
and message_id <= ‘2014-11-25 23:59:59’
and message_id >= ‘2014-11-20 00:00:00’;

Queries
58
Get message by message_id only
Get message by date interval
SELECT * FROM mailbox WHERE message_id = ‘2014-11-21 16:00:00’; ???
SELECT * FROM mailbox WHERE
and message_id <= ‘2014-11-25 23:59:59’ ???
and message_id >= ‘2014-11-20 00:00:00’;

Queries
59
Get message by message_id only (#partition not provided)
Get message by date interval (#partition not provided)
SELECT * FROM mailbox WHERE message_id = ‘2014-11-21 16:00:00’;
SELECT * FROM mailbox WHERE
and message_id <= ‘2014-11-25 23:59:59’
and message_id >= ‘2014-11-20 00:00:00’;

Without #partition
60
No #partition
☞ no token
☞ where are my data ?
C*
C*
C*
C*
C* C*
C* C*
❓ ❓
❓ ❓
❓
❓
❓
❓

Queries
61
Get message by user range (range query on #partition)
Get message by user pattern (non exact match on #partition)
SELECT * FROM mailbox WHERE login >= hsue and login <= jdoe;
SELECT * FROM mailbox WHERE login like ‘%doe%‘;

WHERE clause restrictions
62
All DML queries must provide #partition
Only exact match (=) on #partition, range queries (<, ≤, >, ≥) not allowed
•  ☞ full cluster scan
On clustering columns, only range queries (<, ≤, >, ≥) and exact match (=)
WHERE clause only possible
•  on columns deﬁned in PRIMARY KEY
•  on indexed columns ( )

63
What if I want to perform "arbitrary" WHERE clause ?
•  search form scenario, dynamic search ﬁelds

64
DO NOT RE-INVENT THE WHEEL !
•  ☞ Apache Solr (Lucene) integration (Datastax Enterprise Search)
•  ☞ Same JVM, 1-cluster-2-products (Solr & Cassandra)

65
DO NOT RE-INVENT THE WHEEL !
•  ☞ Apache Solr (Lucene) integration (Datastax Enterprise Search)
•  ☞ Same JVM, 1-cluster-2-products (Solr & Cassandra)
SELECT * FROM users WHERE solr_query = 'age:[33 TO *] AND gender:male';
SELECT * FROM users WHERE solr_query = 'lastname:*schwei?er';

66
Q & A
! "

Advanced Data Modeling
67

Collection types
68
login text,
name text,
age int,
friends set<text>,
hobbies list<text>,
languages map<int, text>,
…

User Deﬁned Type (UDT)
69
Instead of
login text,
…
street_number int,
street_name text,
postcode int,
country text,
…

User Deﬁned Type (UDT)
70
CREATE TYPE address (
street_number int,
street_name text,
postcode int,
country text);
login text,
…
location frozen <address>,
…

UDT Insert
71
INSERT INTO users(login,name, location) VALUES (
'jdoe',
'John DOE',
{
'street_number': 124,
'street_name': 'Congress Avenue',
'postcode': 95054,
'country': ‘USA’
});

JSON syntax for INSERT/UPDATE/DELETE
72
id text PRIMARY KEY,
age int,
state text );
INSERT INTO users JSON '{"id": "user123", "age": 42, "state": "TX"}’;
INSERT INTO users(id, age, state) VALUES('me', fromJson('20'), 'CA');
UPDATE users SET age = fromJson('25’) WHERE id = fromJson('"me"');
DELETE FROM users WHERE id = fromJson('"me"');

JSON syntax for SELECT
73
> SELECT JSON * FROM users WHERE id = 'me';
[json]
----------------------------------------
{"id": "me", "age": 25, "state": "CA”}
> SELECT JSON age,state FROM users WHERE id = 'me';
[json]
----------------------------------------
{"age": 25, "state": "CA"}
> SELECT age, toJson(state) FROM users WHERE id = 'me';
age | system.tojson(state)
-----+----------------------
25 | "CA"

Why Materialized Views ?
Relieve the pain of manual denormalization
74
CREATE TABLE user(
id int PRIMARY KEY,
country text,
…);
CREATE TABLE user_by_country(
country text,
id int,
…,
PRIMARY KEY(country, id));

Materialzed View In Action
75
CREATE MATERIALIZED VIEW user_by_country
AS SELECT country, id, firstname, lastname
FROM user
WHERE country IS NOT NULL AND id IS NOT NULL
PRIMARY KEY(country, id)
CREATE TABLE user_by_country (
country text,
id int,
firstname text,
lastname text,
PRIMARY KEY(country, id));

User Deﬁned Functions (UDF)
76
CREATE [OR REPLACE] FUNCTION [IF NOT EXISTS]
maxOf (col1 int, col2 int)
CALL ON NULL INPUT | RETURNS NULL ON NULL INPUT
RETURN int
LANGUAGE java
AS $$
return Math.max(col1, col2);
$$;
SELECT maxOf(col1, col2) FROM table WHERE id = xxx;

User Deﬁned Aggregates (UDA)
77
CREATE [OR REPLACE] AGGREGATE [IF NOT EXISTS]
sum(bigint)
SFUNC accumulatorFunction
STYPE bigint
[FINALFUNC ﬁnalFunction]
INITCOND 0;
CREATE FUNCTION accumulatorFunction(accu bigint, column bigint)
RETURNS NULL ON NULL INPUT RETURN bigint LANGUAGE java
AS $$ return accu + colum; $$;

78
Q & A
! "

79
@doanduyhai
duy_hai.doan@datastax.com
https://academy.datastax.com/
Thank You

Cassandra introduction 2016

More Related Content

What's hot (20)

Viewers also liked (17)

Similar to Cassandra introduction 2016 (20)

More from Duyhai Doan (8)

Recently uploaded (20)

Cassandra introduction 2016