Redis和Memcached在Ruby上的性能评测（2）——读取性能

liuyuehua

上一篇我们测试了memcached和redis在ruby上的写入性能，这次我们来测试一下它们的单纯读取性能。
测试环境不变，代码只需要进行一点点修改
memcached测试代码：

require 'rubygems'
require 'memcached'
@cache = Memcached.new('localhost:11211')
100000.times do |i|
@cache.get "Test/#{rand(10000)}"
end


redis测试代码：

require 'rubygems'
require 'hiredis'
@cache = Hiredis::Connection.new
@cache.connect("127.0.0.1", 6379)
100000.times do |i|
@cache.write ['GET', "Test/#{rand(10000)}"]
@cache.read
end


测试结果：

realusersys
memcached1x GET pipeline10000 times0.783s0.274s0.285s
redis1x GET pipeline10000 times1.114s0.417s0.375s
memcached10x GET pipeline10000 times1.634s2.969s1.745s
redis10x GET pipeline10000 times3.360s5.204s2.077s
memcached20x GET pipeline10000 times3.163s6.081s3.587s
redis20x GET pipeline10000 times6.588s10.634s4.338s
memcached1x GET pipeline100000 times6.723s1.786s2.658s
redis1x GET pipeline100000 times9.157s1.644s2.109s
memcached10x GET pipeline100000 times12.237s17.363s14.919s
redis10x GET pipeline100000 times24.886s16.058s14.649s
memcached20x GET pipeline100000 times24.184s35.505s29.955s
redis20x GET pipeline100000 times47.903s32.386s29.155s
memcached1x GET pipeline1000000 times1m19.704s15.792s23.092s
redis1x GET pipeline1000000 times1m30.097s15.186s24.438s
memcached10x GET pipeline1000000 times1m57.467s2m44.216s2m23.563s
redis10x GET pipeline1000000 times3m58.644s2m5.219s2m16.882s
memcached20x GET pipeline1000000 times3m51.706s5m32.873s4m48.830s
redis20x GET pipeline1000000 times7m28.133s4m5.430s4m28.178s


测试结果和写入时的基本一致，包括速度和资源占用。由此我们可以得出它们各自的使用场景：
memcached：性能就是一切，很大的并发量和读写量，有专门的服务器作为缓存机，这时候似乎就只能选择memcached了。
而以下情况可以考虑用redis替代：
1、并发量很小或每秒的读写次数比较有限
2、没有单独的cache服务器，还有其他比较消耗cpu的服务在运行
3、内存不是很足够（可以用redis的hash来作替代方案，内存可以节约6/7，参考：http://redis.io/topics/memory-optimization）
4、希望cache复杂数据类型
所以，当只追求性能的时候选择memcached，其他时候都可以尝试使用redis替代～