The document discusses different NoSQL databases and how Cassandra compares to them. It notes that Cassandra uses a Dynamo-inspired architecture with Bigtable-style columns. Cassandra provides better write performance than MySQL through its use of consistent hashing and replication across multiple data centers for high availability. It also offers better read performance than MySQL for large datasets through its use of column-oriented storage.

2. (24)
• @sunsuk7tp
• /P.A. WORKS /
• CS M2
• :
 : HPC
 TSUBAME
 MPI, Cell B.E., GPU CUDA, Hadoop on
 :

 , P2P
 NoSQL Afternoon in Japan (10.11.1, )
 SACSIS 2011
• Web 6
 PHP, Perl, JavaScript

 Apache Solr, MySQL
 NoSQL
 NoSQL
• Jazz, trumpet
• Cassandra 0.6.0
 @railute @yutuki_r @techmemo
Itmedia
3
http://lab.jibun.atmarkit.co.jp/entries/1058

3. +
 NoSQL, Key-Value Store (KVS), Document-Oriented DB, GraphDB
: memcached, Google Bigtable, Amazon Dynamo, Amazon SimpleDB, Apache Cassandra,
Voldemort, Ringo, Vpork, MongoDB, CouchDB, Tokyo Tyrant, Flare, ROMA, kumofs, Kai, Redis,
LevelDB, Hadoop HBase, Hypertable,Yahoo! PNUTS, Scalaris, Dynomite, ThruDB, Neo4j, IBM
ObjectGrid, Oracle Coherence, 100
 : ↔
•
• join, transaction
• /

4. /DC
• decentralized
•
• master/slave
• data/meta/proxy
•
• • Map Reduce
•

5.  SPOF
 DC
dc1 dc2
rack/dc
region
dc3

6. 
•

• ( ) << &
• , correlated failure
 SPOF = “ ”

• : 1
• : /
( ) Daniel Ford et. al. (Google), “Availability in Globally Distributed Storage Systems”, OSDI 2010

7. 
⇒ !!
 SPOF
 ~
SPOF

8.  decentralized
• proxy/master/slave


9. Consistent Hashing ( )

(A~Z )
N := 3 ID
A F
Z • request proxy
secondary 1
• primary node
Q • secondary node
V N
primary secondary 2
hash(key) = Q
key values

10. MyCassandra

11. SQL map
Megastore
library
relational
data model table
(multi-dimentional
sorted map)
(sorted) records (sorted) map
(sorted) map
+ indices + indices
RDB Bigtable KVS
NoSQL

12. PNUTS (VLDB ‘08): MySQL NoSQL YCSB (SOCC ’10):

13. Write-Heavy Read-Heavy
write-
optimized
Better
read- read-
optimized optimized
write-
optimized

14. Apache HBase write optimized Bigtable like centralized
Apache Cassandra write optimized Bigtable like decentralized
Sharded MySQL read optimized MySQL centralized
Yahoo! Sherpa read optimized MySQL centralized
:
⇒ Cassandra MySQL

15. MyCassandra

16. = Dynamo + Bigtable

17. = Dynamo + Bigtable
(P2P/decentralized)

18. = Dynamo +
(P2P/decentralized)

19.  RDBMS
 Table
• /
•
•
NoSQL !!
query

20. MyCassandra

21. = Dynamo +
(P2P/decentralized)

22. MySQL
= Dynamo + Bigtable
Redis
:

24. 1
(master/worker, sharding,
consistent hashing)
• cache / persistence
• index
• write/read-optimized
•

25. +
MyCassandra

26.  InnoDB (MySQL 5.1~ )
 MyISAM
 Memory
 Merge
 Archive
 Federated
 NDB
 CSV
 Blackhole ( )
 FALCON
 MariaDB
 Drizzle InnoDB/MyISAM
 solidDB
MySQL Cluster
 :

27.  MySQL:
 Bigtable: Cassandra
 Redis: / snapshot
 MongoDB: DB



28.  decentralized
•
 RDB (MySQL / PostgreSQL)
• master/slave decentralized
 MongoDB / Redis

• MapReduce
 MySQL Bigtable
 MySQL (InnoDB) INSERT
 Bigtable INSERT/GET
•
 / /
 EC2+RDS MyCassandra

29.  /
 I/O
• Bigtable (LSM-tree)
• MySQL (B-trees/ )
• Redis (Hash)
• MongoDB (B-tree)
• KyotoCabinet (B+ tree/hash)

30. hash B-Trees LSM-Tree
write 1 random I/O append
read 1 random I/O N random I/O
+ merge
cache
Memcached, MySQL, Cassandra,
Redis, MongoDB, HBase,
KyotoCabinet KyotoCabinet LevelDB



31. +
: O(1)
 sequential write
I/O
 Always writable
write-lock memory
sync <k1, obj (v1+v2)> async flush
write path Memtable
LSM-Tree [P. O’Neil ‘96]
disk
<k1, v1>, <k1, v2>
Commit Log
sequential
disk mem <k1,obj1>
write SSTable 1
<k1,obj2>
SSTable 2
<k1,obj3>
SSTable 3
SSTable

32. +
 Key
• Memtable value
• SSTable value
I/O
disk memory
<k1,obj>
Memtable
disk mem disk
<k1,obj+obj1~3>
Commit Log
client merge
<k1,obj1>
SSTable 1
I/O <k1,obj2>
SSTable 2
<k1,obj3>
SSTable 3

33. +
( / 99.9%)
1/9
Better
read write
avg. 6.16 ms
Number of queries
read
Latency (ms)
write write: 2.0 ms
avg. 0.69 ms read: 86.9 ms
99.9 percentile
Latency (ms)

34. Max. QPS for 40 Clients Bigtable
MySQL
40000
Redis
35000
30000
25000
20000
15000
10000
5000 Better
0
(qps) Write Only Write Heavy Read Heavy Read Only

35.  / /
 /99%/Max/


 ( KB~ MB)
 HDD/SSD
 (zipfian, uniform, latest)

• Embedded InnoDB, KyotoCabinet
# ( )

36. select

37. proxy
 client
client
• o.a.c.cli
• o.a.c.avro/thrift server
 proxy
• o.a.c.service.StorageProxy
 server engine
• o.a.c.service.StorageService
• o.a.c.db.ReadVerbHandler/RowMutationVerbHandler
 engine
• o.a.c.db.Table (keyspace )
 o.a.c.db.commitlog
 o.a.c.db.ColumnFamilyStore (columnfamily )
 o.a.c.db.engine.StorageEngineInterface
 o.a.c.db.engine.MySQLInstance, RedisInstance, MongoDBInstance, …

38. 
• put (key, cf)
 OK
• get (key)
• getRangeSlice (startWith, engWith, maxResults)
• truncate/dropTable/dropDB

• secondaryIndex
• expire
• counter (Cassandra-0.8 )

39.  Cassandra
• : keyspace – columnfamily – column
• key/value( )
•
 ColumnFamily SSTable <key, value>
 value: columnFamily
Keyspace
ColumnFamily A ColumnFamily B
key col gender age region key col visits plan
sato male 17 [null] sato 18 Gold
suzuki female 21 Tokyo suzuki 214 Bronze
Bigtable (Cassandra)

40.  Cassandra
• Super Column
 SSTable key-value
•
 KVS
key prefix
•

41. Cassandra MySQL Redis
keyspace database db
column family table record
column field

42. database db
table A table B key values
key values key values
A:sato …
sato gender;male;age;17 sato visits;18;plan;Gold
B:ito …
suzuki gender;female;age; suzuki visits;
A:suzuki …
21;region;Tokyo 214;plan;Bronze
B:tanaka …
RDB (MySQL)
KVS (Redis)
keyspace
columnfamily A columnfamily B
key col gender age region key col visits plan
sato male 17 [null] sato 18 Gold
suzuki female 21 Tokyo suzuki 214 Bronze
Bigtable (Cassandra)

43. 
• MySQL database = keyspace :=>
 MyCassandra (MySQL)
• MySQL table = keyspace :=>
 Cassandra Bigtable (Cassandra)
keyspace
columnfamily A columnfamily B
key col gender age region key col visits plan
sato male 17 [null] sato 18 Gold
suzuki female 21 Tokyo suzuki 214 Bronze
MySQL
gender age region visits plan
sato male 17 [null] 18 Gold
Table
suzuki female 21 Tokyo 214 Bronze

44. 
1

secondary index
rowKey CF counter secondary token
index
Serialized
Object
Key Value
Key-Value KVS …

45. 
•
•
•
write query read query
sync async async sync
W R W R
Bigtable MySQL Bigtable MySQL

46. • W:
• R:
• RW:


write query
sync async
W R

Quorum Protocol: ( )+ ( )> ( )
•
write read
W RW R

47. • :
• R:
• RW:
=3, =2
Client
W:RW:R = 1:1:1 Proxy
1)
2) W, RW
ACK
ACK
3a)
W RW R
3b) R
ACK
: max (W, RW)

48. • :
• R:
• RW:
=3, =2
W:RW:R = 1:1:1 Client
Proxy
1)
2) R, RW
3a)
3b) or
W
W RW R
4)
.
(Cassandra read repair )
: max (R, RW)

49. 20000 Cassandra
0.90 max. qps for 40 clients MyCassandra Cluster
18000
16000 6.49
14000
12000 1.54
0.93
10000
Better 8000
6000
4000
2000
0
[100:0] [50:50] [5:95] [0:100] [write:read]
(query/sec) Write-Only Write-Heavy Read-Heavy Read-Only
Write Heavy Read Heavy
• YCSB / Zipfian
• 6.49
•

50.  https://github.com/sunsuk7tp/MyCassandra
 MyCassandra-0.2.0 ( )
• based on Cassandra-0.7.5
• Baseic CRUD on a simple record
• RangeSlice
• keyspace

51. 1. cassandra.yaml
• engine host, port, …
• default engine
2. ( )
3. MyCassandra (Cassandra )
4. or keyspace,
columnfamily
• engine (keyspace )
• (column family )

52.  Embedded InnoDB
• HailDB: …
• Handler Socket: …
• ExtraDB
• API
 DBM (KyotoCabinet)
• KyotoCassandra/Kyossandra/ ssandra ( )
•
• NoSQL
• QDBM, TC Hash or B+Tree db

53. • /
• hash/B+tree
•
class persistence algorithm lock unit
ProtoHashDB volatile hash whole (rwlock)
ProtoTreeDB red black tree whole (rwlock)
StashDB hash record (rwlock)
CacheDB hash record (mutex)
GrassDB B+ tree page (rwlock)
HashDB persistent hash record (rwlock)
TreeDB B+ tree page (rwlock)
DirDB undefined record (rwlock)
ForestDB B+ tree page (rwlock)

54.  MyCassandra-0.2.2
• secondaryIndex
 MySQL MongoDB
 MyCassandra-0.3.0
• Based on Cassandra-0.8
• Atomic counter
• Brisk (Hadoop + Cassandra)…

55. 1.
2.
3.

56.  Cassandra /expire
• tombstone
• SSTable
• Bigtable like
 MyCassandra Bigtable
•
• expire
•
 1 Table

57. 

instance instance instance
ping detect
engine engine engine instance ?
?
node down ?

58. 
•
 Redis
 MongoDB

•
 key
 Join


59. 
•

• Cassandra-0.6 :
 GC

• Cassandra-0.7, 0.8:




 …

60.  Issue
• https://github.com/sunsuk7tp/MyCassandra/issues
 Twitter
• @MyCassandraJP
• @_MyCassandra # @MyCassandra orz
• @sunsuk7tp #
 Google Groups
• https://groups.google.com/group/my-cassandra

61.  / @railute
• Cassandra
 Gemini Mobile Technologies / @geminimobile
• Hibari
 / @yutuki_r
• Cassandra twitter
 dann / @techmemo
• Cassandra
 / @tatsuya6502
• YCSB , Hibari
 / @mikio1978 / @fallabs
• KyotoCabinet
 / @muga_nishizawa
 / @Nakata_itpro
 / @shudo
 Cassandra

 UST ( )