我正在blitz.io和新遗物上测试一个rails 4.0应用程序..
Blitz: 200 users, 4000 Timeout, in 60sec.
Heroku: 20 dynos (2x standard dynos).
Database: Postgresql 200 connections plan
Using Unicorn at 7 processes(new relic shows that my app uses about 140mb,
so for a 2x standard dyno, you get 1gb, and so 1024/140 is about 7. link ref for this in below line)
ref: http://blog.bignerdranch.com/2188-how-to-improve-rails-application-performance-on-heroku/
对于相同的配置和多个测试，我看到新文件中的响应排队范围从240ms到443ms。我想知道为什么它变化那么多。还有什么我应该监控。怎么能把它减少到至少50ms？什么是理想的响应排队时间？
感谢