差别
这里会显示出您选择的修订版和当前版本之间的差别。
|
toaster:stress-test-report [2015/03/18 15:56] cobub |
toaster:stress-test-report [2017/07/14 11:18] |
||
|---|---|---|---|
| 行 1: | 行 1: | ||
| - | ====== 性能测试报告 ====== | ||
| - | ===== 1. 测试环境如下表所描述： ===== | ||
| - | |||
| - | ^操作系统 ^CentOS-6.4-64bit^ | ||
| - | |CPU |Intel(R) Xeon(R) CPU E5-2620 0 @ 2.00GHz * 12| | ||
| - | |RAM |32 GiB| | ||
| - | |Kernel |2.6.32-358.el6.x86_64| | ||
| - | ===== 2. 按照[[toaster:installation-guide-of-pns|推送服务器安装指南]]文档的说明，安装并配置 Cobub Toaster。 ===== | ||
| - | |||
| - | |||
| - | ===== 3. 切换到安装目录下，假设是/opt/toaster，启动Cobub Toaster： ===== | ||
| - | <code> | ||
| - | cd /opt/toaster | ||
| - | ./toaster -log_dir=”./log” | ||
| - | </code> | ||
| - | ===== 4. 修改并多次启动测试程序，分别建立50000-500000的clients连接，然后用ganglia监控硬件资源，调用推送服务器的WEB接口 /allpub 来测试不同数量 clients 时推送服务器的响应速度。最终得出的数据如下： ===== | ||
| - | |||
| - | |||
| - | ^客户端数量 ^推送耗时(ms) ^推送成功率 ^测试前内存占用(GiB) ^稳定后内存(GiB) ^Toaster内存占用(GiB)^ | ||
| - | |50,000 |582.8 |100.00% |2.7 |4.4 |1.7| | ||
| - | |60,000 |687.0 |100.00% |2.7 |4.0 |1.3| | ||
| - | |70,000 |814.7 |100.00% |3.3 |4.6 |1.3| | ||
| - | |80,000 |933.0 |100.00% |2.7 |4.3 |1.6| | ||
| - | |90,000 |1046.1 |100.00% |2.1 |4.5 |2.4| | ||
| - | |100,000 |1164.4 |100.00% |1.8 |5.2 |3.4| | ||
| - | |200,000 |2320.2 |100.00% |2.7 |6.6 |3.9| | ||
| - | |300,000 |3478.5 |100.00% |4.0 |9.0 |5.0| | ||
| - | |400,000 |4615.5 |100.00% |3.3 |13.4 |10.1| | ||
| - | |500,000 |5851.2 |100.00% |2.6 |12.3 |9.7| | ||
| - | {{:toaster:20150318-075634.png}} | ||
toaster/stress-test-report.txt · 最后更改: 2017/07/14 11:18 (外部编辑)