Naarad is a framework for performance analysis & rating of sharded & stateful services.
Naarad is a highly configurable system analysis tool that parses and plots timeseries data for better visual correlation. It can be used for performance analysis of your service/application. You collect data for the metrics you want to monitor and:
The power of Naarad is in its configurablity. You can use it to glance at various metrics and then choose the important metrics to plot to visually correlate the metrics together. An example use-case is when your application's throughput dropped, you want to know if it was because of some GC activity or a spike in CPU usage or disk I/O. Naarad can help you investigate such issue better.
Many tools and frameworks like Zenoss, rrdtool, etc have solved the use-case of metric collection, parsing and plotting. Naarad has an overlap in functionality with these tools, but the main advantage of naarad is in its flexibility, which lets it be a powerful tool for performance investigations. Naarad users are performance experts who need to look for 'needle in a haystack'. Naarad was built to support this use-case.
git clone https://github.com/linkedin/naarad.git
cd naarad; pip install -r requirements.txt
Please check out our wiki page for more details on Naarad's usage, supported metrics etc.