Gather round, EMC's ol' man Tucci knows Big Data's 'killer app' Joe Tucci, EMC boss and non-Oracle oracle speaking at Oracle OpenWorld, said that predictive real-time analysis would be the killer Big Data application. EMC's big kahuna was key-noting on the second day at the event, following on from Larry Ellison's opening session on the first day, where Ellison gloated over Oracle's …
I work for EMC so I'm biased, but I gotta say that I agree with the "style" comment.. Tucci is an extremely credible leader. Larry is too, from what I hear, but in public, he just comes across as a bit of an ass, which Tucci NEVER does.
AC cause I don't want to be accused of brown-nosing.
I am really not keen on a world where all companies engage in dynamic price manipulation to fork over their customers to the maximum possible extent.
For example, the BA website bait and switch function where you go look for a flight and you see there are loads of seats and the the prices are quite sensible. Somehow though, in the 5 mins between you first looking for the flight and actually trying to buy it the plane is so full you can't use any "BA miles" and a fully restricted economy seat is $15 trillion one-way with no changes or upgrades allowed.
Of course they only do this to you once you are logged in but thanks to the "security" bottom inspectors it's not like you can sign in with another name to get the $200 price again.
Bunch of Utter Bar Stewards.
Of course offering a different price based on paying Google to tell you what post / zip code the victim is in or them using a browser which screams "mark" (Safari?) is entirely ethical. Perhaps then we can call the iPad the "I saw you coming"?
True, they don't understand "dynamic prices" is something customers hate, noone likes to see crazy prices moving continuosly and while trying to understand when to buy something and at what price. They may think they "maximize" revenues, but truly don't understand how many sales they miss because customers who can wait or don't really need an item go away.
Dynamic pricing is pretty much a universal practice in the air-travel industry, and has been for a long time.
It's also probably widely practiced, or at least experimented with, in online retail. Amazon's famous 2000 differential-pricing scandal was a particularly public example, but it would be very surprising if data-mining retailers weren't adjusting prices on the fly.
That said, there's nothing special about the "big data" products for enabling this sort of thing. In particular, there's no reason to believe that doing the analysis in real time would convey a competitive advantage. If you can build a model of customer behavior that's sufficiently robust to increase your profits, it's almost certainly nearly as accurate if it's trained on, say, customer behavior up to the last 24 hours, than if it's trained on customer behavior up to the last millisecond. So your data can be processed in batch. Also, much of that data is structured; you might see a marginal improvement by including unstructured data such as customer reviews, but it probably won't be all that large, and that processing doesn't require any special hardware or software.
In short, I think Tucci's missed the boat on this one. Maybe the largest retailers could squeeze enough additional advantage out of EMC's (or anyone else's) proprietary tech to make it worthwhile - but they're also the shops that will just build their own. If you have a smaller operation and you really want to mine user reviews for sentiment, say, someone could throw that together with some commodity hardware and open-source software (eg UIMA, OpenNLP, a little glue code, and something to digest the output).
I was going to complain that "NOSQL" should be "NoSQL", but a little research shows that some people do indeed use it as an acronym[1], for "Not Only SQL". Nasty, but there it is. (At least the "NoSQL" variant seems more common, thank goodness.)
On the other hand, Hadoop is not a "data storage model". It's an implementation of a distributed filesystem, a sharding NoSQL database (technically Hbase, but often lumped in with Hadoop proper), a work partitioning engine, and some other bits. The "data storage model", such as it is, is distributed filesystem and sharding database. Calling Hadoop a data storage model is like calling Microsoft Word a document-authoring model.
[1] None of your "acronym" versus "initialism" nonsense, please. Where people got the idea that English words have to be pronounced phonetically I do not know.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
