[kea-dev] GSoC 2019 with ISC - initial proposal and thoughts

[Ambauen Daniel (ID NET)]

Hello Franek

I’m only a kea user but I would like to share some thoughts about this topic.

1. I would only implement counters no time-series, because our days there are so many powerful 
tools available for collecting time-series (prometheus, Influxdb …).
If you setup kea in in a larger setup, you would like to use a in you department already
present monitoring and alerting system.   

2. Kea should be open to suche monitoring system.
For example a build-in HTTP endpoint for prometheus would be very convenient.  

3. If the API call to retrieve the statistics has an option to reset the counters, 
we could easy retrieve time-series via the API with out additional computation.   
 
4. If kea could write the statistics to a file or to syslog could be interesting as well
for some user.

I wish you look at GSoC 2019

Regards
Daniel



> On 23 Feb 2019, at 09:36, Franciszek Górski <fagorski9 at gmail.com> wrote:
> 
> Hello Dev Team!
> 
> My name is Franek Górski, I am an IT student from Gdańsk University of Technology.
> Some of you might remember me from the Kea Hackathon in Gdańsk which took place in last October.
> Now I would like to take part in GSoC 2019 as a student intern and I am writing to present my thoughts about Idea no 2. titled ‘Statistics enhancements’ from yours Ideas for Kea Community Projects.
> I would be very grateful for more infromation about this project and response whether my ideas are correct including tips what should I change in my assumptions, what can I add.
> 
> My initial thoughts and suggestions about the problem:
> 
> 1. The primary task is to expand the method of collecting observations for implementing time-series ‘buckets’. Now for each statistic we only keep one value by overwriting the previous one.
> 
> 2. If we want to collect more than one value of our statistics in a storage we should change the method of setting/adding value. Now it’s simple overwriting of the first element which is the only element, in a list of samples’s storage. I propose changing it to a vector consisting the same pairs of values – time stamp of observation and the value of that observation. New pair should be add at the begging of the vector to simplify the access to the current value.
> 
> 3. I don’t understand exactly the idea of how much data we would like to collect about such statistic, but it seems to be that we cannot only adding new values without setting some limits of maximal size of storage. I was considering about using not yet implemented functions in Stats_Mgr like setMaxSampleAge and setMaxSampleCount for appointing some thresholds.
> 
> 4. Additional functions for managing statistics could be added to Stats_Mgr as member functions or like an inner class or they might be included in a separate class but implemented in the same files.
> 
> I still have some other considerations about but the main problem for me is lack of more informations concerning the project.
> 
> 
> Regards
> Franek
> _______________________________________________
> kea-dev mailing list
> kea-dev at lists.isc.org
> https://lists.isc.org/mailman/listinfo/kea-dev