Rack Aggregate

Statistics aggregator middleware based on Joseph Ruscio's aggregate gem which provides: minimum, maximum, mean, standard deviation measures of your applications response latency. Best of all, it accomplishes this without recording or storing any of the actual sample values - meaning, by itself it will not impact your performance or memory footprint even across millions of requests.

Example

> ruby examples/rackapp.rb
> ab -c1 -n20 http://127.0.0.1:4567/ (run 20 requests against the app)
> ...
> curl http://127.0.0.1:4567/aggregate

count:    20
mean:    4100.30ms
min:    1000.11ms
max:    9000.17ms
std_dev:  2447.42ms

Request histogram (buckets in ms):
value |------------------------------------------------------------------| count
  512 |@                                                                 |     1
 1024 |@@@@@@@                                                           |     7
 2048 |@@@@                                                              |     4
 4096 |@@@@@@                                                            |     6
 8192 |@@                                                                |     2
      ~
Total |------------------------------------------------------------------|    20

Quickstart with Rack

gem install rack-aggregate

# in your rack app / rackup file
require 'rack/aggregate'
use Rack::Aggregate

Quickstart with Rails 3

# in your Gemfile
gem 'rack-aggregate', :require => 'rack/aggregate'

# in development.rb environment
config.middleware.use Rack::Aggregate

Wishlist

• Configurable aggregate path
• Custom sampling rate - no need to log every request
• Other output formats? JSON?

License

(The MIT License) Copyright © 2011 Ilya Grigorik