MySQL Cluster performance test results and doubt
I redhat as5 4G mem environment inside revealed a cluster environment.
4 Data node, 4 Taiwan MySQL node, 1-Taiwan mgm node, 2 Taiwan do LVS front load balance
Test result:
insert the standard mysql operation and down almost 80 percent
select and standard operating mysql down almost 50 percent
LVS little impact on performance, but can play a role in load-sharing
I use the 5.0 belt inside the cluster, depends entirely on memory, not written disk. Note that I think the performance bottlenecks before network bandwidth to this, I just click carefully, client to mysql api node bandwidth occupied very low, unless it is mysql node and data node bandwidth communications between the bottleneck.
There is a question:
Why four data node there is a master? Did not find the document describes, but only master, is able to write, all able to read the model »
In addition, Tanzi Li of people have used cluster on the production » Can discussion, I was considering whether we will move to the cluster of data.
Mysql see this year’s annual meeting that the future will consider the combination of memcached and mysql, there is no experiment had » Feeling and Sina mentioned in the memcachedb and dbcached almost concept.
[color = Red] [size = 6] add some of the CGE test Goodies: [/ size] [/ color]
[attach] 251821 [/ attach]
[attach] 251822 [/ attach]
[root @ localhost ~] # ndb_mgm
– NDB Cluster - Management Client –
ndb_mgm> show
Connected to Management Server at: 192.168.1.230:1186
Cluster Configuration
———————
[ndbd (NDB)] 4 node (s)
id = 2@192.168.1.230 (mysql-5.1.24 ndb-6.3.14, Nodegroup: 0, Master)
id = 3@192.168.1.232 (mysql-5.1.24 ndb-6.3.14, Nodegroup: 0)
id = 4@192.168.1.234 (mysql-5.1.24 ndb-6.3.14, Nodegroup: 1)
id = 5@192.168.1.236 (mysql-5.1.24 ndb-6.3.14, Nodegroup: 1)
[ndb_mgmd (MGM)] 1 node (s)
id = 1@192.168.1.230 (mysql-5.1.24 ndb-6.3.14)
[mysqld (API)] 4 node (s)
id = 6@192.168.1.230 (mysql-5.1.24 ndb-6.3.14)
id = 7@192.168.1.232 (mysql-5.1.24 ndb-6.3.14)
id = 8@192.168.1.234 (mysql-5.1.24 ndb-6.3.14)
id = 9@192.168.1.236 (mysql-5.1.24 ndb-6.3.14)
ndb_mgm>
[color = Red] Yesterday, a simple test version of the CGE, have no pre-LVS circumstances, insert performance has not improved, select and enhance performance, there are probably increase 20-25 percent. Note to the non-index are not open to write-disk functionality test cases
300w records, a single record of 1 K, for the test results, not LVS circumstances, performance is almost the MySQL Server of 70%, LVS front with two MySQL nodes, essentially equivalent to MySQL Server, LVS mysql with four points, performance ratio MySQL Server, and connect with a few more, the more obvious [/ color]
Later, the trial of the ndbapi,’ll read and write, the same environment than the number of mysql server even better, of course, is the mem circumstances. Most are written row disk cases, and almost mysqlserver
