**TL;DR: ** if you have outgrown Crossfilter and need something that behaves like it, but scales to tenths of millions aggregates, this is for you.
This is Crossfilter:
It is an incredibely usefull small library and probably one of the best ways of digging into simple data invented so far. I believe that every organisation that requires at least some amount of data analysis would benefit from it. And therefore, it is an obligation for every data techie to somehow demonstrate it to colleagues.
There is a beautiful visualisation library built on top of it, have a look, it is called dc.js.
In fact, if your aggregate count will never exceed a million of rows, you probably just should stop here, read no further and enjoy the simplicity of the current client-side solution.
Imagine this hipothetical situation where you adopt a Crossfilter based solution in your org and it suddenly picks up. Actually, it is not that hypothetical - this happened every time so far in my personal history.
What happens if you are lucky enough to work in one of those companies or departments that experience so called exponential growth? What happens, if your data grows massively?
I'l tell you, what. You are screwed with your beautiful interactive graphs alltogether.
Implementing a data analysis solution that relies on browser capacity only is a bit like defecating in your own pants: you get an instant sensation of relief and warmness, but than it somehow start to smell funny. And everyone notices it very quickly.
First of all, retrieving a couple of million records takes time. Processing those rows might make your browser unresponsive. If you are using json to load data, you might easily hit the hardcoded limit of a maximum json string that can be deserialized in V8. You can handle these cases with pagination, but believe me, there is a good reason, why it is set so. Browsers struggle to process more.
Second, you most probably will run out of 8, 16 dimensions, and you might even exceed the hard limit of 32 dimensions supported by Crossfilter.
Third, you probably will have to start managing the csv/json fileds containing the data. Adding/removing fields and old data has to be managed somehow, and the larger is the dataset, the more effort it takes to do so.
All this will quickly become a major point of pain. This is why this project has been created.
This project is a server-side simulation of Crossfilter. Only the final aggregates are downloaded to the client, and every time you click on a filter, the request is sent to the server side, processed, and new results are retrieved.
Because of this architecture, it is quite hard to achieve 30ms reaction speed, however, with powerful servers and fast networks still allow you to achieve speeds <100ms even for larger datasets.