Cassandra Community Webinar - Introduction To Apache Cassandra 1.2

CASSANDRA COMMUNITY WEBINARS APRIL 2013
INTRODUCTION TO
APACHE CASSANDRA 1.2
Aaron Morton
Apache Cassandra Committer, Data Stax MVP for Apache Cassandra
@aaronmorton
www.thelastpickle.com
Licensed under a Creative Commons Attribution-NonCommercial 3.0 New Zealand License

Cassandra Summit 2013
June 11 & 12
San Francisco
Use SFSummit25 for 25% off

Cassandra Summit 2013
DataStax Ac*ademy
Free certiﬁcation during the
summit.

Overview
The Cluster
The Node
The Data Model

Cassandra
- Started at Facebook
- Open sourced in 2008
- Top Level Apache project
since 2010.

Used by...
Netﬂix,Twitter, Reddit,
Rackspace...

Inspiration
- Google Big Table (2006)
- Amazon Dynamo (2007)

Why Cassandra?
- Scale
- Operations
- Data Model

Store ‘foo’ key with Replication Factor 3.
Node 1 - 'foo'
Node 2 - 'foo'Node 4
Node 3 - 'foo'

Consistent Hashing.
- Evenly map keys to nodes
- Minimise key movements
when nodes join or leave

Partitioner.
RandomPartitioner
transforms Keys to Tokens
using MD5.
(Default pre version 1.2.)

Partitioner.
Murmur3Partitioner
transforms Keys to Tokens
using Murmur3.
(Default in version 1.2.)

Keys andTokens?
token 0 99
key 'fop' 'foo'
10 90

Token Ring.
'foo'
token: 90
'fop'
token: 10
99 0

Token Ranges pre v1.2.
Node 1
token: 0
Node 2
token: 25
Node 4
token: 75
Node 3
token: 50
1-2576-0

Token Ranges withVirtual Nodes in v1.2.
Node 1
Node 2
Node 3
Node 4

LocateToken Range.
Node 1
Node 2
Node 3
Node 4
'foo'
token: 90

Replication Strategy selects
Replication Factor number of
nodes for a row.

SimpleStrategy with RF 3.
Node 1
Node 2
Node 3
Node 4
'foo'
token: 90

NetworkTopologyStrategy uses
a Replication Factor per Data
Centre.
(Default.)

Multi DC Replication with RF 3 and RF 2.
Node 1
Node 2
Node 3
Node 4
'foo'
token: 90
Node 1
Node 2
Node 3
Node 4
West DC East DC

The Snitch knows which Data
Centre and Rack the Node is
in.

SimpleSnitch.
Places all nodes in the same
DC and Rack.
(Default, there are others.)

EC2Snitch.
DC is set to AWS Region and
a Rack to Availability Zone.

The Client and the Coordinator.
Node 1
Node 2
Node 3
Node 4
'foo'
token: 90
Client

Multi DC Client and the Coordinator.
Node 1
Node 2
Node 3
Node 4
'foo'
token: 90
Client
Node 10
Node 20
Node 30
Node 40

Gossip.
Nodes share information with
a small number of neighbours.
Who share information with a
small number of neigh..

Consistency Level (CL).
- Speciﬁed for each request
- Number of nodes to wait
for.

Consistency Level (CL)
- Any*
- One,Two Three
- QUORUM
- LOCAL_QUORUM, EACH_QUOURM*

QUOURM at Replication Factor...
Replication
Factor
QUOURM
2 or 3 4 or 5 6 or 7
2 3 4

Write ‘foo’ at QUOURM with Hinted Handoff.
Node 1
Node 2
Node 3
Node 4
'foo' for #3
'foo'
token: 90
Client

Read ‘foo’ at QUOURM.
Node 1
Node 2
Node 3
Node 4
'foo'
token: 90
Client

Column Timestamps
used to resolve
differences.

Resolving differences.
Column Node 1 Node 2 Node 3
purple
cromulent
(timestamp 10)
cromulent
(timestamp 10)
<missing>
monkey
embiggens
(timestamp 10)
embiggens
(timestamp 10)
debigulator
(timestamp 5)
dishwasher
tomato
(timestamp 10)
tomato
(timestamp 10)
tomacco
(timestamp 15)

Consistent read for ‘foo’ at QUOURM.
Node 1
Node 2
Node 3
Node 4
Client
cromulent
cromulent
<empty>
Node 1
Node 2
Node 3
Node 4
Client
cromulent cromulent

Strong Consistency
W + R > N
(#Write Nodes + #Read Nodes> Replication Factor)

Achieving Strong Consistency.
- QUOURM Read + QUORUM Write
- ALL Read + ONE Write
- ONE Read + ALL Write

Achieving Consistency
- Consistency Level
- Hinted Handoff
- Read Repair
- Anti Entropy

Write path
Append to Write
Ahead Log.
(fsync every 10s by default, other options available)

Write path...
Merge Columns
into Memtable.
(Lock free, always in memory.)

(Later.)
Asynchronously ﬂush
Memtable to new ﬁles.
(May be 10’s or 100’s of MB in size.)

Data is stored in
immutable SSTables.
(Sorted String table.)

SSTable ﬁles.
*-Data.db
*-Index.db
*-Filter.db
(And others)

SSTables.
SSTable 1
foo:
dishwasher (ts 10):
tomato
purple (ts 10):
cromulent
SSTable 2
foo:
frink (ts 20):
ﬂayven
monkey (ts 10):
embiggins
SSTable 3 SSTable 4
foo:
dishwasher (ts 15):
tomacco
SSTable 5

Read purple, monkey, dishwasher.
SSTable 1-Data.db
foo:
dishwasher (ts 10):
tomato
purple (ts 10):
cromulent
SSTable 2-Data.db
foo:
frink (ts 20):
ﬂayven
monkey (ts 10):
embiggins
SSTable 3-Data.db SSTable 4-Data.db
foo:
dishwasher (ts 15):
tomacco
SSTable 5-Data.db
Bloom Filter
Index Sample
SSTable 1-Index.db
Bloom Filter
Index Sample
SSTable 2-Index.db
Bloom Filter
Index Sample
SSTable 3-Index.db
Bloom Filter
Index Sample
SSTable 4-Index.db
Bloom Filter
Index Sample
SSTable 5-Index.db
Memory
Disk

Key Cache caches row key
position in *-Data.db ﬁle.
(Removes up to1disk seek per SSTable.)

Read with Key Cache.
SSTable 1-Data.db
foo:
dishwasher (ts 10):
tomato
purple (ts 10):
cromulent
SSTable 2-Data.db
foo:
frink (ts 20):
ﬂayven
monkey (ts 10):
embiggins
foo:
dishwasher (ts 15):
tomacco
SSTable 5-Data.db
Key Cache
Index Sample
SSTable 1-Index.db
Key Cache
Index Sample
SSTable 2-Index.db
Key Cache
Index Sample
SSTable 3-Index.db
Key Cache
Index Sample
SSTable 4-Index.db
Key Cache
Index Sample
SSTable 5-Index.db
Memory
Disk
Bloom Filter Bloom Filter Bloom Filter Bloom Filter Bloom Filter

Row Cache caches entire row.
(Removes all disk IO.)

Read with Row Cache.
Row Cache
SSTable 1-Data.db
foo:
dishwasher (ts 10):
tomato
purple (ts 10):
cromulent
SSTable 2-Data.db
foo:
frink (ts 20):
ﬂayven
monkey (ts 10):
embiggins
foo:
dishwasher (ts 15):
tomacco
SSTable 5-Data.db
Key Cache
Index Sample
SSTable 1-Index.db
Key Cache
Index Sample
SSTable 2-Index.db
Key Cache
Index Sample
SSTable 3-Index.db
Key Cache
Index Sample
SSTable 4-Index.db
Key Cache
Index Sample
SSTable 5-Index.db
Memory
Disk
Bloom Filter Bloom Filter Bloom Filter Bloom Filter Bloom Filter

Compaction merges truth from
multiple SSTables into one
SSTable with the same truth.
(Manual and continuous background process.)

Compaction.
Column SSTable 1 SSTable 2 SSTable 4 New
purple
cromulent
(timestamp 10)
<tombstone>
(timestamp 15)
<tombstone>
(timestamp 15)
monkey
embiggens
(timestamp 10)
embiggens
(timestamp 10)
dishwasher
tomato
(timestamp 10)
tomacco
(timestamp 15)
tomacco
(timestamp 15)

Cassandra is good at
reading data from a row in the
order it is stored.

Typically an efﬁcient data model will
denormalize data and use the
storage engine order.

To create a good data model
understand the queries your
application requires.

API Choice
Thrift
Original and still fully
supported API.

API Choice
CQL3
New and fully supported API.

CQL 3
A Table Orientated, Schema
Driven, Data Model and
Query language similar to
SQL.

Twitter clone
using CQL 3 via the cqlsh
tool.
bin/cqlsh

Queries?
- Post Tweet to Followers
- Get Tweet by ID
- List Tweets by User
- List Tweets in User Timeline
- List Followers

Keyspace
A Namespace container.

Our Keyspace
CREATE KEYSPACE
cass_community
WITH replication =
{'class':'NetworkTopologyStrategy',
'datacenter1':1};

Table
A sparse collection of well
known, ordered columns.

FirstTable
CREATE TABLE User
(
user_name text,
password text,
real_name text,
PRIMARY KEY (user_name)
);

Some users
cqlsh:cass_community> INSERT INTO User
... (user_name, password, real_name)
... VALUES
... ('fred', 'sekr8t', 'Mr Foo');
cqlsh:cass_community> select * from User;
user_name | password | real_name
-----------+----------+-----------
fred | sekr8t | Mr Foo

Some users
cqlsh:cass_community> INSERT INTO User
... (user_name, password)
... VALUES
... ('bob', 'pwd');
cqlsh:cass_community> select * from User where user_name =
'bob';
user_name | password | real_name
-----------+----------+-----------
bob | pwd | null

Data Model (so far)
Table /
Value
User
user_name Primary Key

TweetTable
CREATE TABLE Tweet
(
tweet_id bigint,
body text,
user_name text,
timestamp timestamp,
PRIMARY KEY (tweet_id)
);

TweetTable...
cqlsh:cass_community> INSERT INTO Tweet
... (tweet_id, body, user_name, timestamp)
... VALUES
... (1, 'The Tweet','fred',1352150816917);
cqlsh:cass_community> select * from Tweet where tweet_id = 1;
tweet_id | body | timestamp | user_name
----------+-----------+--------------------------+-----------
1 | The Tweet | 2012-11-06 10:26:56+1300 | fred

Data Model (so far)
Table /
Value
User Tweet
user_name Primary Key Field
tweet_id Primary Key

UserTweetsTable
CREATE TABLE UserTweets
(
tweet_id bigint,
user_name text,
body text,
PRIMARY KEY (user_name, tweet_id)
);

UserTweetsTable...
cqlsh:cass_community> INSERT INTO UserTweets
... VALUES
... (1, 'The Tweet','fred',1352150816917);
cqlsh:cass_community> select * from UserTweets where
user_name='fred';
user_name | tweet_id | body | timestamp
-----------+----------+-----------+--------------------------
fred | 1 | The Tweet | 2012-11-06 10:26:56+1300

UserTweetsTable...
cqlsh:cass_community> select * from UserTweets where
user_name='fred' and tweet_id=1;
-----------+----------+-----------+--------------------------
fred | 1 | The Tweet | 2012-11-06 10:26:56+1300

UserTweetsTable...
cqlsh:cass_community> INSERT INTO UserTweets
... VALUES
... (2, 'Second Tweet', 'fred', 1352150816918);
cqlsh:cass_community> select * from UserTweets where user_name = 'fred';
-----------+----------+--------------+--------------------------
fred | 1 | The Tweet | 2012-11-06 10:26:56+1300
fred | 2 | Second Tweet | 2012-11-06 10:26:56+1300

UserTweetsTable...
cqlsh:cass_community> select * from UserTweets where user_name = 'fred' order by
tweet_id desc;
-----------+----------+--------------+--------------------------
fred | 2 | Second Tweet | 2012-11-06 10:26:56+1300
fred | 1 | The Tweet | 2012-11-06 10:26:56+1300

UserTimeline
CREATE TABLE UserTimeline
(
user_name text,
tweet_id bigint,
tweet_user text,
body text,
PRIMARY KEY (user_name, tweet_id)
)
WITH CLUSTERING ORDER BY (tweet_id DESC);

UserTimeline
cqlsh:cass_community> INSERT INTO UserTimeline
... (user_name, tweet_id, tweet_user, body, timestamp)
... VALUES
... ('fred', 1, 'fred', 'The Tweet',1352150816917);
cqlsh:cass_community> INSERT INTO UserTimeline
... (user_name, tweet_id, tweet_user, body, timestamp)
... VALUES
... ('fred', 100, 'bob', 'My Tweet',1352150846917);

Data Model (so far)
Table /
Value
User Tweet
User
Tweets
User
Timeline
user_name Primary Key Field Primary Key Primary Key
tweet_id Primary Key
Primary Key
Component
Primary Key
Component

UserMetricsTable
CREATE TABLE UserMetrics
(
user_name text,
tweets counter,
followers counter,
following counter,
PRIMARY KEY (user_name)
);

UserMetricsTable...
cqlsh:cass_community> UPDATE
... UserMetrics
... SET
... tweets = tweets + 1
... WHERE
... user_name = 'fred';
cqlsh:cass_community> select * from UserMetrics where
user_name = 'fred';
user_name | followers | following | tweets
-----------+-----------+-----------+--------
fred | null | null | 1

Data Model (so far)
Table /
Value
User Tweet
User
Tweets
User
Timeline
User Metrics
user_name
Primary
Key
Field
Primary
Key
Primary
Key
Primary
Key
tweet_id
Primary
Key
Primary Key
Component
Primary Key
Component

Relationships
CREATE TABLE Followers
(
user_name text,
follower text,
PRIMARY KEY (user_name, follower)
);
CREATE TABLE Following
(
user_name text,
following text,
PRIMARY KEY (user_name, following)
);

Relationships
cqlsh:cass_community> INSERT INTO
... Following
... (user_name, following, timestamp)
... VALUES
... ('bob', 'fred', 1352247749161);
cqlsh:cass_community> INSERT INTO
... Followers
... (user_name, follower, timestamp)
... VALUES
... ('fred', 'bob', 1352247749161);

Data Model
Table /
Value
User Tweet
User
Tweets
User
Timeline
User
Metrics
Follows
Followers
user_name
Primary
Key
Field
Primary
Key
Primary
Key
Primary
Key
Primary
Key
tweet_id
Primary
Key
Primary Key
Component
Primary Key
Component

Aaron Morton
@aaronmorton
www.thelastpickle.com
Licensed under a Creative Commons Attribution-NonCommercial 3.0 New Zealand License

Cassandra Community Webinar - Introduction To Apache Cassandra 1.2

More Related Content

What's hot (20)

Viewers also liked (11)

Similar to Cassandra Community Webinar - Introduction To Apache Cassandra 1.2 (20)

More from aaronmorton (19)

Recently uploaded (20)

Cassandra Community Webinar - Introduction To Apache Cassandra 1.2