Re: {Spam?} RE: Something all DBAs should see!

Ian Michael Gumby wrote: > > > From: obnoxio@serendipita.com > > > > Ian Michael Gumby wrote: > > > > http://www.cloudera.com/blog/2009/11/11/hadoop-world-rethinking-the-data-warehouse-with-hadoop-and-hive-from-ashish-thusoo/ > > > > > > You may not believe in Hadoop/HBase, but here's a good presentation on > > > building a world class large scale data warehouse. > > > > That's nice, Gumby, but let's just look at this: they're not using ANY > > database. They're not even using Hadoop. They are building their very > > own database specifically to underpin their business. I suspect that > > Facebook is also full of programmers who like nothing more than creating > > their own infrastructure, so they didn't investigate commercial options > > very thoroughly. > > > Huh? > No truer words spoken from a clown. > > Let me think.... > The speaker is talking at a Hadoop conference and he's talking about > using Hive. > Hive just happens to be a sub project under Apache's Hadoop. Yeah, so they're using completely unmodified Hadoop, are they? No splits? No "contributions"? Oh ... ! > > This will, I predict, bite them severely in the arse, if it hasn't > > already. But you'll never hear about it. It would fuck their IPO or > > share price. > > > LOL. > > Spoken like a true dinosaur. > I mean, sure if Facebook was worried, they wouldn't talk about it. > But what about Amazon, Yahoo!, Google, and others? > Oh Look! There's an IBM logo somewhere up there. Yeah, because none of them use bespoke software to underpin their business, do they? Fucking hell, Gumby, Amazon are so far out there that they've given as a completely different computing paradigm. It's called the cloud. They just had the money and the resource to struggle through all the complexities and make it work. But very few businesses need that much of an out-there thing and have the money so that they are prepared to go through the pain of inventing a completely new paradigm. > > What you are proposing would inevitably have consequences for all the > > other users of IDS. And I can promise you that for all the sizzle of > > column-oriented data, there is one killer problem: *no-one will buy it*. > > You may have it bundled into a few ELA's and some sad life donor like > > yourself may actually unwrap the software and fiddle with it, but no > > company is actually going to use it and it's certainly NOT going to sell > > enough to recoup the investment. > > > > Ah yes, now we get to it. > > Back in '97 Informix had a chance to get on Linux well ahead of the > competition. > Know what they said? "If you can build a business case, we'll listen." > Well we built a business case. > Did they listen? > Nope. > And had they listened, they would have learned that the then CEO of Red > Hat was talking to Larry's folks so they knew everything Informix said. > Why else do you think they knew the release date so they could trump it? 8^0 Well, that doesn't exactly bolster your case, does it? I firmly believed from the outset that Linux would be a game changer. I entirely agree that Informix screwed the pooch with the Linux port. And having said all those things *I* think that this port would be a complete waste of time and money. I'm not speaking for IBM here, or second-guessing them. *I* don't think anyone would buy a Hadoop DataBlade, and I'd far rather IBM spent the money they'd spend on a Hadoop DataBlade on a benchmark or marketing. (Wishful thinking, I know!) > So I think that the bozo's who run IBM's IM team are likely to do the > same thing. > I mean heck, there's a couple of WW job titles that are being filled by > complete and utterly incompetent morons, so why pay attention to someone > who's always been ahead of the curve? So far ahead of the curve, he's up his own arse, you mean? :o) > Security? Naw that's not important. But then again, where's the morons > who were on the i.Sell team back in '98? > Yeah that was a real brain trust. > > Sun getting a bid from Oracle? Naw Oracle's a software company. They > would never want a hardware company. What, exactly, does this have to do with Informix? What, for that matter, does the i.Sell "brains trust" have to do with IBM? You seem determined to tie together every mistake or misjudgment or thing that you disagreed with, ever, by anyone related in any way, shape or form to Informix. > > And here's the other thing you need: > > http://hadoop.apache.org/hbase/releases.html > > > > Not necessarily. > HBase sits on Hadoop. You don't need it, in fact you (IBM) don't want it. > But that's a discussion over your pay grade. ;-) > Also you don't want to do HDFS but respond to the APIs. Big difference. Oh well. It's two APIs then. Easy for a genius like you to tie them together, surely? Trivial, even. > > Put your money where you mouth is, Gumby: write the Hadoop/HBase > > DataBlade. Then go sell it and show me how fucking wrong I was. > > > Because you clown, suppose I invest the time and energy. What's my ROI > when IBM can and will release the same thing behind me? Who'll buy my They will? You really think that they wouldn't just buy you out? > datablade when IBM can either sell it for the same price but offer tons > more support and guarantee future compatibility in sync with an engine > release. Have they *ever* done this? > How many people are using the user written RoR adaptor? I mean since IBM > released their DRDA RoR Adaptor? Lots, I suspect, since I hear that the IBM adapter is clunky and badly written. > Of course that would mean that people are using Ruby over Grails which > requires a simple JDBC interface. ;-) > (But the lesson was already learned.) And I will now proceed to pleasure myself with this fish. > Yeah, unlike some clowns, I do my homework and determine the level of > effort, and the tROI and ROI I could expect. And what ROI can IBM expect for this? If you go to Jerry and say "here's a case for $x million", do you think he'll do it? So what ROI *can* IBM expect? Because if it's bigger than the development effort, I can't see why it wouldn't happen. Show us your homework then. > Not to mention the fact that were *I* to put HBase on top of IDS, I > would be in fact facing litigation from IBM. Remember my post mentioning > a certain patent a couple of weeks back? ;-) Hint... what's HDFS? Gumby, you talk SO much shit that I actively work to forget your posts. > And of course, if I did a benchmark and go to publish the results, how > long do you think it would take before IBM's lawyers were all over my > ass? Hmmm. yeah thought so. Who knows? Who cares? > And then there's the cost of implementation and the gamble that if you > can build it, you can show a large enough cost savings by not having to > scale to 1000's of nodes to offset your licensing costs. (Oh yes, I > thought about that too. ;-) ) Ah! So it's a "gamble" then? Not a guaranteed ROI? I thought you said you'd done all your homework and thoug