Multithread And Epoll Server To Revolve Online (Maestro)

This project aims to create a tiny but functional application server by leveraging multithread and epoll features of linux OS.

All the source code is written in C. Please go to https://github.com/grassroot72/Maestro2 for better code structure

The project is a WIP (Work In Progress) project. It is still in its infant stage. More features will be added in the future.

Features

• Multithread pool
• Epoll (so, linux specific)
• Non-blocking
• HTTP/1.1 GET method (static file)
• HTTP/1.1 HEAD method (static file)
• HTTP/1.1 POST method
• HTTP/1.1 chunked transfer
• HTTP/1.1 keep-alive (long connection, disconnected after timeouts)
• built-in cache to provide better GET performance
• deflate compression
• download resumption

Cloning

git clone https://github.com/grassroot72/Maestro.git

Build & Run

$ make
...
$ ./maestro

Test

After you run the webserver, you should see the following

listening on port [9000]

then, you can open firefox or other browsers and key in the url http://localhost:9000 or http://localhost:9000/bench.html

To test download resumption, key in the following url in your browser(firefox), http://localhost:9000/download/Business-English-for-Success.pdf refresh the browser if download stalls. Keep refresh several times until the download finally finishes.

Benchmark

For Apache Bench, I used the following command,
$ ab -k -n 500000 -c 1000 http://localhost:9000/bench.html

You may also want to put the compression option on,
$ ab -k -n 500000 -c 1000 -H "Accept-Encoding: deflate" http://localhost:9000/bench.html

Please replace localhost with your hostname if you would like to test on
other computers.

The following is the test result on my very low-end laptop,

• CPU: Intel Celeron N2930 (4 cores) at ~2.16 GHz (2M Cache)
• MEM: 4G
• NET: laptop built-in wireless

This is ApacheBench, Version 2.3 <$Revision: 1879490 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/

Benchmarking localhost (be patient)
Completed 50000 requests
Completed 100000 requests
Completed 150000 requests
Completed 200000 requests
Completed 250000 requests
Completed 300000 requests
Completed 350000 requests
Completed 400000 requests
Completed 450000 requests
Completed 500000 requests
Finished 500000 requests


Server Software:        Maestro/1.0
Server Hostname:        localhost
Server Port:            9000

Document Path:          /bench.html
Document Length:        1025 bytes

Concurrency Level:      1000
Time taken for tests:   34.448 seconds
Complete requests:      500000
Failed requests:        0
Keep-Alive requests:    500000
Total transferred:      669500000 bytes
HTML transferred:       512500000 bytes
Requests per second:    14514.75 [#/sec] (mean)
Time per request:       68.895 [ms] (mean)
Time per request:       0.069 [ms] (mean, across all concurrent requests)
Transfer rate:          18979.73 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0    0   3.6      0     105
Processing:    27   69  18.4     65     178
Waiting:        5   43  13.6     45     137
Total:         27   69  18.7     65     181

Percentage of the requests served within a certain time (ms)
  50%     65
  66%     73
  75%     79
  80%     84
  90%     96
  95%    105
  98%    115
  99%    122
 100%    181 (longest request)

If you test it on most recently manufactured computers, you will get much
better benchmark results.

Note

I am not the author of the pdf file under the download folder, I only use it for download resumption test.

Todo

I want to develop an HTTP/1.1 compatible webserver as well as an application server which skips traditional CGI function and directly provides dynamic services (POST) through built-in functions.

todo list:

• HTTPS