ndhist

[jovo]

can we discuss the pro's and con's of leaving the 'stats' module within ndstore, vs. making it its own repo?

making it its own repo:

pro's

• keeps things more modular
• is an example "plug-in" that in theory, other people could contribute as well
• it makes the ndpaper and probably the ndviz/ndtilecache papers a bit cleaner.

  con's

• it does not work on its own (but this is true of all plug-ins)
• it will take some effort to move it.

@kunallillaney @randalburns @alexbaden thoughts?

[alexbaden]

I'm fine with the idea.

There is a snag, which is the histograms are stored in the ocpuser / nduser models file. So, we would have to decouple the histograms from the backend metadata. But the only advantage to the coupling, at the moment anyway, is that deleting a channel will also delete its associated histogram (in theory, anyway -- in fact, this may not be true and if so would be a bug).

I also suspect the timeline for doing this is > 1 month. Even just porting the stats service to a django app inside ndstore is work. And if we're going to separate it, there might be some advantages to running it as a more standalone micro-service.

[jovo]

oh > 1 month is a lot. ok, let's punt until after writing the papers :) eg, we could write a conference paper later on micro-services, with this being the prime example.