I don't have much time to write tonight, but I did come up with a potentially interesting course correction in response to the conversation with my Set Theory prof. Actually, it's motivated by more than just that one conversation (I'm not that easily swayed; actually, I'm not easily swayed at all). It also occurred to me during a conversation on Friday with some engineers from IBM why nobody seems that interested in my topic. It's good ideas, but my basic premise is wrong. Data is not expanding faster than hardware capabilities. Hardware is actually doing a pretty good job of keeping up. It only seems like data is expanding faster because companies don't like having to replace their hardware every year. However, they don't much like replacing their software, either. So, if it's going to be a switch, they'll probably just buy the bigger box and leave the software alone.
That's not to say that there's nothing to be gained via software. There certainly is. And, that's the course correction. Rather than force a 1-off solution (that I can't possibly implement as well as the full teams at HP, IBM, Oracle, and others), I'm going to go up a layer and look more at the motivation of these problems. My goal is still data reduction, but it's a little more along traditional lines of how to best boil down data rather than how to best avoid looking at it at all.
More refinement to come as I work this through.
No comments:
Post a Comment