Please don't be a jerk in response to someone sharing what they're working on. If people can't share that without being jumped on, this quickly becomes a shitty place for conversation. Probably you didn't mean it that way, but it's difficult to gauge intent on the internet, so the burden is on the commenter to dismabiguate it.
I have used all of the existing solutions that can scale to the workloads I require. None of them are adequate, all have deficiencies either in cost, scalability, ergonomics or operability.
If you read the rest of my comments here it should be clear this is something I have thought about for some time and have a well reasoned architecture and completely different take on how this problem should be approached.
This is not about features, this is about fundamentally changing the storage architecture from layout of the log data itself, to the indices to the query engine, even the distributed system model.
Most of the ideas are stolen from battle tested systems like Druid which I have worked with extensively and other systems I respect like Pilosa, Ok Log (which was a project in a similar vein).
Have a little optimism, things can be better if we just sit down and make them so.
