Merge pull request #171 from Mashape/document/benchmark

Document/benchmark

README.md

@@ -7,7 +7,7 @@
 - Mailing List: [Google Groups](https://groups.google.com/forum/#!forum/konglayer)
 - Gitter Chat: [mashape/kong](https://gitter.im/Mashape/kong)
 
-**Kong is next-generation software for securing, managing and extending APIs & Microservices**. If you're building for web, mobile or IoT you will likely end up needing to implement common functionality on top of your actual software. Kong can help by acting as a gateway for any HTTP resource while providing logging, authentication and other functionality through plugins. 
+**Kong was built for securing, managing and extending APIs & Microservices**. If you're building for web, mobile or IoT you will likely end up needing to implement common functionality on top of your actual software. Kong can help by acting as a gateway for any HTTP resource while providing logging, authentication and other functionality through plugins. 
 
 Powered by NGINX and Cassandra with a focus on high performance and reliability, Kong runs in production at [Mashape](https://www.mashape.com) where it has handled billions of API requests for over ten thousand APIs.
 
@@ -38,6 +38,34 @@ Full versioned documentation is available at [GetKong.org](http://getkong.org):
 - [API Reference](http://getkong.org/docs/latest/admin-api)
 - [Configuration](http://getkong.org/docs/latest/configuration/)
 
+## Benchmarks
+
+We set Kong up on AWS and load tested it to get some performance metrics. The setup consisted of three `m3.medium` EC2 instances; one for Kong, another running Cassandra and a third acting as the target API powered by Lwan. All servers were tuned with the following:
+
+1) Added `fs.file-max=80000` to `/etc/sysctl.conf`
+
+2) Added the following lines to: `/etc/security/limits.conf`
+```
+*          soft     nproc          80000
+*          hard     nproc          80000
+*          soft     nofile         80000
+*          hard     nofile         80000
+```
+
+We then load tested the Kong server ramping up from 100 to 1000 concurrent connections. The benchmark generating 65,084 requests over 120 seconds with an average of 542 req/second or about 46,860,480 req/day. 
+
+#### Response Data:
+
+- Success: 100.00% 
+- Error: 0.00% 
+- Timeout: 0.00% 
+
+- Fastest Response: 4 ms
+- Slowest Response: 31 ms
+- Average Response: 6 ms
+
+![](http://i.imgur.com/ROxu4AP.png)
+
 ## Development
 
 Please see the [CONTRIBUTING.md][kong-contrib] if you would like to have your changes merged into Kong.