选择vire-benchmark而不是redis-benckmark，是因为vire-benchmark支持多并发压测。
vire-benchmark 是vire里面带的工具
源码仓库：https://github.com/vipshop/vire/tree/master
编译的方法参考git上就行，make后，会生成 vire 文件，此外在 tests目录下还有3个二进制文件（ viretest、 vireabtest 、vire-benchmark）生成。
下面是我在本机二进制或者docker-compose方式启动的3个kv实例：
redis7 端口 6379 （本机二进制部署）
keydb6.3.0 端口 6389 （本机docker部署）
dragonfly latest 端口 6499 （本机docker部署）
下面只压测了最常见的几种操作指令
（因为压测工具和被压实例都部署在一台机器上，所以下面的指标仅供参考，正式压测请分开独立部署）
代码语言：javascript复制压测redis7
$ ./vire-benchmark -h 192.168.31.181 -p 6379 -n 100000 -c 100 -T 16 -t set,get,hset,hget | egrep "====|requests"
====== SET ======
100000 requests completed in 2.49 seconds
40080.16 requests per second
====== GET ======
100000 requests completed in 2.41 seconds
41476.57 requests per second
====== HSET ======
100000 requests completed in 2.62 seconds
38109.75 requests per second
====== HGET ======
100000 requests completed in 2.60 seconds
38491.14 requests per second
压测keydb6.3
$ ./vire-benchmark -h 192.168.31.181 -p 6389 -n 100000 -c 100 -T 16 -t set,get,hset,hget | egrep "====|requests"
====== SET ======
100000 requests completed in 1.67 seconds
59952.04 requests per second
====== GET ======
100000 requests completed in 1.62 seconds
61538.46 requests per second
====== HSET ======
100000 requests completed in 1.76 seconds
56947.61 requests per second
====== HGET ======
100000 requests completed in 1.68 seconds
59488.40 requests per second
压测dragonfly
$ ./vire-benchmark -h 192.168.31.181 -p 6499 -n 100000 -c 100 -T 16 -t set,get,hset,hget | egrep "====|requests"
====== SET ======
100000 requests completed in 0.93 seconds
107296.14 requests per second
====== GET ======
100000 requests completed in 0.86 seconds
116279.07 requests per second
====== HSET ======
100000 requests completed in 0.94 seconds
106837.61 requests per second
====== HGET ======
100000 requests completed in 0.87 seconds
115606.94 requests per second可以看到dragonfly的性能最强，主要是因为它原生的多线程， keydb也是多线程模式（keydb 6.3.0 在我压测过程中出现内存bug自动重启了）。而这里redis测出来最弱是因为单机的，对此redis官方也发过博客提到这个测试的不公平性。
