Closed pocha closed 11 years ago
Let me write a script to benchmark every def in the reel server.
Its not just defs. I have a feeling that searching for the user in the hash could be time consuming too if the hash has too many users. But yes - get started first.
@pocha Pushed in initial code for the ruby-profiler. Have a look and let me know.
The profiling looks good. Can you please add where the values would be dumped.
Also - is it possible to include the delay from the intermediate Rails app in the dump.
On 24-Apr-2013, at 8:24 PM, Saurav wrote:
@pocha Pushed in initial code for the ruby-profiler. Have a look and let me know.
? Reply to this email directly or view it on GitHub.
the profile values with get dumped inside "reel_server_profiler.log" file. The file will be created inside session/sample/ directory, once the code gets executed.
@pocha I will look into the details about what you had mentioned earlier in the mail and get back to you by tomorrow.
Sounds good @csaurav . Append details in README as profiling data. That will help gathering feedback from other experts as to where to improve the performance.
I have removed @vireshas from the issue/milestone. I will assign the issue to the person who is able to put down the profiling data (preferably end to end)
@vireshas, as discussed yesterday, have pushed in the profiler log file for your reference e28dbdd
@pocha, I will add the details in the README soon, bit tied up now.
The output a command takes a lot of time to appear on the output. Its best that the code (reel-server.rb) should be profiled & time taken by various modules should be dumped in a log file to figure out where the latency is.
To test latency in real-time, siege can be used to send some real-time back to back inputs for different users. Analyzing the log file should point to as what thing took maximum time & then that bottleneck should be worked on.