Re: {Spam?} The future of databases?
Obnoxio The Clown
> Ok,
>
> Here are two articles that came up in a google search that happen to
> have caught my eye.
> http://www.networkworld.com/news/2009/121709-big-three-database-vendors-diverge.html?page=1
> http://www.itworldcanada.com/news/big-three-database-vendors-disagree-on-hadoop/139613-pg2
>
> If they look similar, its because both articles are from the same author.
>
> The point, for those who are reading challenged, is that unlike the
> clown, Hadoop and HBase have some merit and that they've caught the eyes
> of some senior management types. IBM, Microsoft and Oracle should be
> concerned because of this on several different levels.
I never said HBase has no merit. I said it has no merit to the people
who'd buy IDS. Cisco buys a lot of IDS, but they haven't bought one
great-big-fuck-off system, they've bought hundreds of little systems (is
what I understand from the blurb, anyway). Same with K-mart or whoever
else buys a lot of Informix. They don't buy great-big-fuck-off single
Informix systems, they buy hundreds or thousands of small systems. I
don't know all Informix's customers, but I've been to hundreds and I've
never seen any of them big enough to justify HBase.
What possible use is HBase to them?
And even though Informix has a great BI architecture, it's not seen as a
BI platform, so even if it was reasonable to integrate HBase somehow AND
it actually worked, someone would have to start selling IDS as a BI
platform.
Which would mean going round to all the DB2 customers and saying "Erm..."
Which would mean going round to all the tools vendors and saying "Erm..."
Etc.
It's a huge gamble with no clear customer target. People who need HBase
are already looking at HBase, they may like the idea of a nice SQL
interface, but by the time the HBase DataBlade does everything people
need, we'll be on the Next Big Thing. People who don't need it aren't
going to be impressed by an HBase DataBlade.
I've been in the IT industry long enough to know when I'm looking at the
Next Big Thing. And HBase isn't it. It's a Thing. It may even be a Big
Thing. But it's not the Next Big Thing.
> Anant D. Jhingran from IBM's IM team seemed to be interested in this....
>
> Now he's the CTO of IBM SWG IM pillar, no?
I wonder who decided to implement PureXML? I wonder how much of that
investment has been recouped by people actually buying PureXML licenses?
> I wonder who some of those select customers are... ;-)
> Maybe one of them OEM's IDS?
So you think they might implement a flaky as fuck oddball product with a
completely different paradigm instead of a reliable, well-known, clearly
understood product in their highly reliable solutions?
> But hey! What do I know?
I don't know. What *do* you know?
--
Cheers,
Obnoxio The Clown
http://obotheclown.blogspot.com
I will now proceed to pleasure myself with this fish.
--
This message has been scanned for viruses and
dangerous content by OpenProtect(http://www.openprotect.com), and is
believed to be clean.
Ian Michael Gumby
> Date: Thu, 17 Dec 2009 22:21:02 +0000
> From: obn...@serendipita.com
> Subject: Re: {Spam?} The future of databases?
> CC: inform...@iiug.org
>
> Ian Michael Gumby wrote:
> > Ok,
> >
> > Here are two articles that came up in a google search that happen to
> > have caught my eye.
> > http://www.networkworld.com/news/2009/121709-big-three-database-vendors-diverge.html?page=1
> > http://www.itworldcanada.com/news/big-three-database-vendors-disagree-on-hadoop/139613-pg2
> >
> > If they look similar, its because both articles are from the same author.
> >
> > The point, for those who are reading challenged, is that unlike the
> > clown, Hadoop and HBase have some merit and that they've caught the eyes
> > of some senior management types. IBM, Microsoft and Oracle should be
> > concerned because of this on several different levels.
>
> I never said HBase has no merit. I said it has no merit to the people
> who'd buy IDS. Cisco buys a lot of IDS, but they haven't bought one
> great-big-fuck-off system, they've bought hundreds of little systems (is
> what I understand from the blurb, anyway). Same with K-mart or whoever
> else buys a lot of Informix. They don't buy great-big-fuck-off single
> Informix systems, they buy hundreds or thousands of small systems. I
> don't know all Informix's customers, but I've been to hundreds and I've
> never seen any of them big enough to justify HBase.
>
I've seen startups w less money than a Hyatt, Peapod, K-Sears, TransUnion, etc ... using Hadoop/Hive/HBase.
As to cost...
Answer me this...
Dual Socket, quad core Intel server (8 cores total). How much to put IDS EE on that machine as a single node? No discounting, MFSR price.
For under 5K, you can build the following in a 1U or 2U high box...
2 socket, 8 core Intel server (5500 family current generation), 4 1TB 3.5" SATA drives and 32GB of memory.
10 Nodes is a small 'cloud'. That's 50K.
So how big do you have to be for you to afford that?
BTW, if you don't care about putting them in a rack, you could build them in PCs, go with just i7 chips (4 cores) and DDR2 memory. Custom built, you could build this for ~$800.00 (usd) a node. (Actually less)
Granted this doesn't remove the database altogether. And its not a solution where a general purpose database is a good fit. But then again, you can feed a relational engine like Postgress off this configuration. (oops. I meant to say IDS. Now why would I mention Postgress? ;-)
But in DW work where you have lots of data, its got its merits. ;-)
Remember that First Union 27TB of data from the 90's?
That's an application where this would fit. (Ooops! That was XPS, my bad. I didn't mean to embarrass IBM for dis'ing the better warehouse machine in favor of DB2... ;-)
> What possible use is HBase to them?
>
> And even though Informix has a great BI architecture, it's not seen as a
> BI platform, so even if it was reasonable to integrate HBase somehow AND
> it actually worked, someone would have to start selling IDS as a BI
> platform.
But yes, you are correct. IDS isn't a BI platform. It is however a general purpose RDBMs that has a very well designed framework that allows for extensibility. So designing something to sit on IDS would be interesting. But that's a discussion that will not take place in public. ;-)
We'll just have to see.
-G
Hotmail: Trusted email with powerful SPAM protection. Sign up now.