对于一个能够支撑超高并发的大型分布式系统来说，像Redis这类分布式缓存是必不可少。Redis在单机部署的模式下，QPS几乎不可能超过10万+，除非机器配置特别好且Redis操作不太复杂。
我们知道，对于数据库来说，如果想要提升读写性能，最简单的方式就是做 一主多从+读写分离 。对于分布式缓存也是一样的道理，因为缓存一般都是用来支撑读请求的高并发，写请求相对较少（一般也就每秒一两千写请求），所以非常适合读写分离的架构。
关于Redis的复制原理，我在进阶篇的《分布式框架之高性能：Redis主从同步》已经详细讲解过了，不熟悉的读者可以先去了解下。
一、主从架构搭建
在生产环境下，我们必须要将Master节点的持久化功能打开，否则万一Master宕机后重启，此时Slave连上Master后，会触发一次全量复制，master就会将空的数据集同步到slave上去，导致Slave中的数据也被清空。
我们先来搭建一个一主二从的Redis构架，我首先在ressmix-dsf02这个节点上安装单机版本的Redis。具体的安装步骤不再赘述，读者可以参考我在《Redis持久化实战》中讲解的搭建步骤。
1.1 配置步骤
搭建完Redis节点后，我们按照以下步骤进行读写分离的配置，ressmix-dsf01作为Master，ressmix-dsf02和ressmix-dsf03作为Slave。（我这里只操作ressmix-dsf02，ressmix-dsf03读者可以自行操作）
- 修改Slave节点的配置文件，配置
replicaof ressmix-dsf01 6379，这样ressmix-dsf01节点就作为了ressmix-dsf02的Master节点； - 强制读写分离：修改Slave节点的配置文件，配置
replica-read-only yes，这样Slave节点会拒绝所有的写操作（Redis 2.6以后Slave节点默认就是只读的，所以这个版本以后的Redis默认可以不设置）； - 集群安全认证：修改Slave节点的配置文件，配置
masterauth ressmix，其中ressmix是我设置的认证密码； - 停止Master节点，然后修改Master节点的配置文件，配置
requirepass ressmix； - 主从节点均配置
appendonly yes，开启AOF持久化； - 绑定节点IP：修改Slave节点的配置文件，配置
bind 192.168.0.109，其中192.168.0.109为ressmix-dsf02的IP，同理也把Master节点的这个配置修改下。另外，为了以防外一，每个节点都执行下iptables -A INPUT -ptcp --dport 6379 -j ACCEPT，用于放开6379端口，然后清理下防火墙：sudo iptables -F。
上述操作全部配置完成后，我们通过以下命令启动主节点，然后以相同方式启动从节点ressmix-dsf02。
cd /etc/init.d
./redis_6379 start
节点启动后，我们可以先在Master节点中写入一条记录：
redis-cli -h 192.168.0.107 -a ressmix
set k1 v1
然后在ressmix-dsf02节点可以查看到同步过来的数据：
我们可以通过执行info replication查看主从复制的状态：
二、性能压测
搭建完Redis的主从架构后，可以对其做一个基准压测，测一下Redis的性能和QPS。Redis自身提供了redis-benchmark压测工具，可以用于一些简单场景下性能测试。
压测工具位于redis安装包的src目录下：
./redis-benchmark -h 192.168.0.107
常用参数如下：
-c <clients> Number of parallel connections (default 50)
-n <requests> Total number of requests (default 100000)
-d <size> Data size of SET/GET value in bytes (default 2)
我们可以根据自己系统高峰期的业务量来设置参数，比如在高峰期，瞬时最大用户量会达到10万，总请求数为1000万，每条数据的大小为50字节，则可以像下面这样模拟请求：
./redis-benchmark -h 192.168.0.107 -c 100000 -n 10000000 -d 50
压测的结果可能像下面这样，显示了不同操作的每秒请求数：
====== PING_INLINE ======
100000 requests completed in 1.28 seconds
50 parallel clients
3 bytes payload
keep alive: 1
99.78% <= 1 milliseconds
99.93% <= 2 milliseconds
99.97% <= 3 milliseconds
100.00% <= 3 milliseconds
78308.54 requests per second
====== PING_BULK ======
100000 requests completed in 1.30 seconds
50 parallel clients
3 bytes payload
keep alive: 1
99.87% <= 1 milliseconds
100.00% <= 1 milliseconds
76804.91 requests per second
====== SET ======
100000 requests completed in 2.50 seconds
50 parallel clients
3 bytes payload
keep alive: 1
5.95% <= 1 milliseconds
99.63% <= 2 milliseconds
99.93% <= 3 milliseconds
99.99% <= 4 milliseconds
100.00% <= 4 milliseconds
40032.03 requests per second
====== GET ======
100000 requests completed in 1.30 seconds
50 parallel clients
3 bytes payload
keep alive: 1
99.73% <= 1 milliseconds
100.00% <= 2 milliseconds
100.00% <= 2 milliseconds
76628.35 requests per second
====== INCR ======
100000 requests completed in 1.90 seconds
50 parallel clients
3 bytes payload
keep alive: 1
80.92% <= 1 milliseconds
99.81% <= 2 milliseconds
99.95% <= 3 milliseconds
99.96% <= 4 milliseconds
99.97% <= 5 milliseconds
100.00% <= 6 milliseconds
52548.61 requests per second
====== LPUSH ======
100000 requests completed in 2.58 seconds
50 parallel clients
3 bytes payload
keep alive: 1
3.76% <= 1 milliseconds
99.61% <= 2 milliseconds
99.93% <= 3 milliseconds
100.00% <= 3 milliseconds
38684.72 requests per second
====== RPUSH ======
100000 requests completed in 2.47 seconds
50 parallel clients
3 bytes payload
keep alive: 1
6.87% <= 1 milliseconds
99.69% <= 2 milliseconds
99.87% <= 3 milliseconds
99.99% <= 4 milliseconds
100.00% <= 4 milliseconds
40469.45 requests per second
====== LPOP ======
100000 requests completed in 2.26 seconds
50 parallel clients
3 bytes payload
keep alive: 1
28.39% <= 1 milliseconds
99.83% <= 2 milliseconds
100.00% <= 2 milliseconds
44306.60 requests per second
====== RPOP ======
100000 requests completed in 2.18 seconds
50 parallel clients
3 bytes payload
keep alive: 1
36.08% <= 1 milliseconds
99.75% <= 2 milliseconds
100.00% <= 2 milliseconds
45871.56 requests per second
====== SADD ======
100000 requests completed in 1.23 seconds
50 parallel clients
3 bytes payload
keep alive: 1
99.94% <= 1 milliseconds
100.00% <= 2 milliseconds
100.00% <= 2 milliseconds
81168.83 requests per second
====== SPOP ======
100000 requests completed in 1.28 seconds
50 parallel clients
3 bytes payload
keep alive: 1
99.80% <= 1 milliseconds
99.96% <= 2 milliseconds
99.96% <= 3 milliseconds
99.97% <= 5 milliseconds
100.00% <= 5 milliseconds
78369.91 requests per second
====== LPUSH (needed to benchmark LRANGE) ======
100000 requests completed in 2.47 seconds
50 parallel clients
3 bytes payload
keep alive: 1
15.29% <= 1 milliseconds
99.64% <= 2 milliseconds
99.94% <= 3 milliseconds
100.00% <= 3 milliseconds
40420.37 requests per second
====== LRANGE_100 (first 100 elements) ======
100000 requests completed in 3.69 seconds
50 parallel clients
3 bytes payload
keep alive: 1
30.86% <= 1 milliseconds
96.99% <= 2 milliseconds
99.94% <= 3 milliseconds
99.99% <= 4 milliseconds
100.00% <= 4 milliseconds
27085.59 requests per second
====== LRANGE_300 (first 300 elements) ======
100000 requests completed in 10.22 seconds
50 parallel clients
3 bytes payload
keep alive: 1
0.03% <= 1 milliseconds
5.90% <= 2 milliseconds
90.68% <= 3 milliseconds
95.46% <= 4 milliseconds
97.67% <= 5 milliseconds
99.12% <= 6 milliseconds
99.98% <= 7 milliseconds
100.00% <= 7 milliseconds
9784.74 requests per second
====== LRANGE_500 (first 450 elements) ======
100000 requests completed in 14.71 seconds
50 parallel clients
3 bytes payload
keep alive: 1
0.00% <= 1 milliseconds
0.07% <= 2 milliseconds
1.59% <= 3 milliseconds
89.26% <= 4 milliseconds
97.90% <= 5 milliseconds
99.24% <= 6 milliseconds
99.73% <= 7 milliseconds
99.89% <= 8 milliseconds
99.96% <= 9 milliseconds
99.99% <= 10 milliseconds
100.00% <= 10 milliseconds
6799.48 requests per second
====== LRANGE_600 (first 600 elements) ======
100000 requests completed in 18.56 seconds
50 parallel clients
3 bytes payload
keep alive: 1
0.00% <= 2 milliseconds
0.23% <= 3 milliseconds
1.75% <= 4 milliseconds
91.17% <= 5 milliseconds
98.16% <= 6 milliseconds
99.04% <= 7 milliseconds
99.83% <= 8 milliseconds
99.95% <= 9 milliseconds
99.98% <= 10 milliseconds
100.00% <= 10 milliseconds
5387.35 requests per second
====== MSET (10 keys) ======
100000 requests completed in 4.02 seconds
50 parallel clients
3 bytes payload
keep alive: 1
0.01% <= 1 milliseconds
53.22% <= 2 milliseconds
99.12% <= 3 milliseconds
99.55% <= 4 milliseconds
99.70% <= 5 milliseconds
99.90% <= 6 milliseconds
99.95% <= 7 milliseconds
100.00% <= 8 milliseconds
24869.44 requests per second
三、总结
本章，我重点讲解了如何进行生产环境的Redis读写分离部署，读者可以自己尝试在虚拟机中动手进行节点部署，以加深印象。
