Professional Documents
Culture Documents
MySQL Cluster
Cluster Interconnect
• Response times
• Throughput
• Low variation of response times
Experience Base
Super
Rest Socket
Read/Write
MySQL Cluster RT solution on
Quad-Core computer using 4 data nodes
CPU optimized architecture
using Dolphin SuperSockets and Polling-based
Commit
Primary Primary
Prepare Commit
Backup Backup
Prepare Commit
Backup Backup
Prepare Commit
• Lock Memory
• Batching of IN (..) primary key access
• INSERT batching
• Condition pushdown (faster table scans)
Lock Memory in Main Memory
• Ensure no swapping occurs in NDB
Kernel
Batching IN (…) with
primary keys
• 100 x SELECT * from t WHERE pk = x;
• SELECT * from t WHERE pk IN
(x1,,,x100);
• IN-statement is around 10x faster than
100 SELECT single record PK access
Use of multi-INSERT
• 100 x INSERT INTO t (x)
• INSERT INTO t (x1),(x2),,,,,(x100)
• Multi-insert up to about 10x faster
Use of features in
MySQL Cluster CGE version 6.4
2 Mysql servers
30000
25000
20000
eth
15000 sci
eth + rt
10000 sci + rt
5000
0
1 2 4 8 16 32
Parallell activity
Distribution aware
(8 data nodes on 2 Quad Core)
120 000
100 000
80 000
eth
60 000 sci
40 000 eth+rt
sci + rt
20 000
0
1 2 4 8 16 32 64 128 256
Parallell activity
%
100
80
60 Improvment sci vs eth
Improvment eth+rt vs eth
40 Improvment sci+rt vs eth
20
0
1 2 4 8 16 32 64 128 256
-20
Parallell activity
Non-distribution aware
(4 data nodes on 4 Quad Cores)
70000
60000
50000
40000
eth
30000
sci
20000
10000
0
1 2 4 8 16 32 64 128 256
Parallell activity