Introducing Druid: Real-Time Analytics at a Billion Rows Per Second
A general overview of Druid and the choices that led up to it.
The next post is said to have details about the architecture, etc.
From what I read here, the holding of all data in memory is one critical part of the solution.
That and having data that can be held in smallish cells.
Tossing blobs, ASCII or binary, into cells, might cause a problem.
Won’t know until the software is available for use by a diverse audience.
I mention it here as an example of defining data sets and requirements in such a way that scalable architectures can be developed, for that particular set of requirements.
There is nothing wrong with having a solution that works best for a particular application.
Ballpoint pens are wonderful writing devices but fail miserably as hammers.
A software or technology solutions that works for your problem is far more valuable than software that solves the general case but not yours.