论坛徽章:: 0

电梯直达

1楼 [收藏(0)] [报告]

发表于 2008-09-04 17:33 |只看该作者 |倒序浏览

看了官方的HA方案觉得DRBD+replication是比较可靠的，参考了精华里面的一篇文章，里面说到不少好东东，但是没有涉及到性能比较，不知道有没有同学和我一样对这种方案的性能很是不满。不知道是不是我哪里配置不对？

硬件环境：
Dell 2950 500G×2 RAID0 其中创建了一个VolGroup单独给DRBD用，4G mem Redhat as5
这样的机器两台

软件环境：
MySQL：5.1.26-rc-community
drbd：version: 8.2.6
HB：2.1.3

配置：
drbd.conf
global
{
      usage-count no;
}

common
{
      protocol C;
}

resource r0
{
      device /dev/drbd0;
      disk    /dev/mapper/VolGroup00-LogVol02;
      meta-disk internal;
      syncer
      {
            rate 100M;
      }
      on drbd-db1
      {
            address 1.1.0.1:7789;
      }
      on db2-raid0
      {
            address 1.1.0.2:7789;
      }
}

my.cnf
[mysqld]
datadir=/drbd/mysql
pid-file=/drbd/mysql/mysql.pid
log-bin
binlog-do-db=mysqlslap    # input the database which should be replicated
binlog-ignore-db=mysql       # input the database that should be ignored for replication
binlog-ignore-db=test
server-id=1
default-storage-engine=InnoDB
character-set-server=utf8
collation-server=utf8_general_ci
innodb_file_per_table
max_connections=1000

[mysql.server]
user=mysql

其中1.1.0.1和1.1.0.2是网线直连

测试数据：
全部使用的是InnoDB引擎
DRBD+MySQL：
mysqlslap -a --concurrency=50,100,200 --number-of-queries 10000 --debug-info --auto-generate-sql-load-type=write
Benchmark
      Average number of seconds to run all queries: 27.835 seconds
      Minimum number of seconds to run all queries: 27.835 seconds
      Maximum number of seconds to run all queries: 27.835 seconds
      Number of clients running queries: 50
      Average number of queries per client: 200

Benchmark
      Average number of seconds to run all queries: 27.719 seconds
      Minimum number of seconds to run all queries: 27.719 seconds
      Maximum number of seconds to run all queries: 27.719 seconds
      Number of clients running queries: 100
      Average number of queries per client: 100

Benchmark
      Average number of seconds to run all queries: 27.088 seconds
      Minimum number of seconds to run all queries: 27.088 seconds
      Maximum number of seconds to run all queries: 27.088 seconds
      Number of clients running queries: 200
      Average number of queries per client: 50

User time 0.17, System time 0.47
Maximum resident set size 0, Integral resident set size 0
Non-physical pagefaults 2717, Physical pagefaults 0, Swaps 0
Blocks in 0 out 0, Messages in 0 out 0, Signals 0
Voluntary context switches 62779, Involuntary context switches 565

MySQL replication：
mysqlslap -a --concurrency=50,100,200 --number-of-queries 10000 --debug-info --auto-generate-sql-load-type=write
Benchmark
      Average number of seconds to run all queries: 0.679 seconds
      Minimum number of seconds to run all queries: 0.679 seconds
      Maximum number of seconds to run all queries: 0.679 seconds
      Number of clients running queries: 50
      Average number of queries per client: 200

Benchmark
      Average number of seconds to run all queries: 0.917 seconds
      Minimum number of seconds to run all queries: 0.917 seconds
      Maximum number of seconds to run all queries: 0.917 seconds
      Number of clients running queries: 100
      Average number of queries per client: 100

Benchmark
      Average number of seconds to run all queries: 1.164 seconds
      Minimum number of seconds to run all queries: 1.164 seconds
      Maximum number of seconds to run all queries: 1.164 seconds
      Number of clients running queries: 200
      Average number of queries per client: 50

User time 0.13, System time 0.52
Maximum resident set size 0, Integral resident set size 0
Non-physical pagefaults 2653, Physical pagefaults 0, Swaps 0
Blocks in 0 out 0, Messages in 0 out 0, Signals 0
Voluntary context switches 60294, Involuntary context switches 5993

以上可以看出写性能似乎差别很大，大家看看有问题吗？欢迎给出建议和你的测试数据

读性能的比较似乎意义不大，应为mysqlslap自动生成的表没有index，而且select全是整个表数据的查询就是类似select × from t1之类的。不过估计性能也会差别比较大吧。