July 24, 2008

How will Oracle save its data warehouse business?

By acquiring DATAllegro, Microsoft has seriously leapfrogged Oracle in data warehouse technology. All doubts about maturity and versatility notwithstanding, DATAllegro has a 10X or better size advantage (actually, I think it’s more like 20-40X) versus Oracle in warehouses its technology can straightforwardly handle. Oracle cannot afford to let this move go unanswered.

It’s of course possible that Oracle has been successfully developing comparable data warehouse technology internally. But it’s unlikely. Oracle hasn’t done anything that radical, internally and successfully, for about 15 years, RAC (Real Application Clusters) excepted. (I.e., since the object/relational extensibility framework started in Release 7.) So in all likelihood, the answer will come via acquisition. I think there are four candidates that make the most sense: Teradata, Vertica, ParAccel, and Greenplum. Kognitio (controlled by former Oracle honcho Geoff Squire) might be in the mix as well. Netezza is probably a non-starter because of its hardware-centric strategy.

Here’s why I’m emphasizing Teradata, Vertica, ParAccel, and Greenplum:

Teradata is the long-time market leader in high-end data warehousing. It would be no embarrassment for Oracle to, in connection with an acquisition, admit this. It also has the most mature and versatile product of any of the alternatives. Teradata is close to being a software-only solution; i.e., it’s not hardware-centric enough to be a problem. While Teradata would be expensive, Oracle could afford it. However, there are tax issues blocking an acquisition too soon after the 2007 spin-out; I don’t recall exactly how long it is before an acquisition would be realistic.

Other than Teradata, Vertica would offer the most marketing pizzazz. Database legend Mike Stonebraker and lesser legend/former Oracle honcho Jerry Held have both declared that Vertica is the future, because it’s columnar. Saying “Yes, those guys were right” would be a pretty straightforward and effective marketing strategy. What’s more, it would be possible to pitch columnar as trumping Microsoft/DATAllegro’s (and Netezza’s) better-row-store offering.

ParAccel also offers the advantages of columnar. Its “Amigo” mode lets it play nicely with OLTP DBMS. And since it originally focused on the SQL Server market and hasn’t won a lot of customers, ParAccel might come relatively cheap.

Greenplum’s technology is sort of an MPPization of a general-purpose relational DBMS (PostgreSQL). Porting that to Oracle would be much harder than DATAllegro’s Ingres-to-SQL-Server porting task, but it should be doable.

Related links

Comments

15 Responses to “How will Oracle save its data warehouse business?”

  1. DW Consultant on July 25th, 2008 6:23 am

    How can porting from Ingres to SQL Server be easier. DatAllegro is all Java based with Infiniband Stack. I donot think MS has done the due dilligence. It might not be a good buy for MS.

  2. Curt Monash on July 25th, 2008 11:16 am

    I’m quite sure Microsoft has done a lot of investigation, and specifically enough to rise to the level of due diligence.

    And yes, all acquisitions involve a risk of failure.

    Your point?

    CAM

  3. Pera Detlic on July 25th, 2008 4:33 pm

    Could IBM snatch NZ next? Would such acquisition make sense?

  4. Niko on July 26th, 2008 3:26 am

    @DW Consultant

    Microsoft did not buy DATAllegro for its technology, it was bought for the name and the appearance of a presence in the data warehouse market. If anyone talks a good game, it is Stuart Frost, the master of spin. The guy could sell ice cubes to Eskimos. And people like Curt (no disrespect) have fallen for his game.

    The entire software stack will be replaced. The underlying database will switch from Ingres to SQL Server, the OS will shift from Linux to Windows Server, and most likely the top level query optimizer/master node software written in Java will likely be rewritten in C/C++ (I think it is hard to imagine Microsoft being a Java shop and the APIs to SQL Server are probably not supported in Java, though that may be a Version 2 swap-out). This basically takes would-have-been DATAllegro customers, off the market for the better part of a year I would say. It is very unfortunate for the half a dozen or so current production customers as the current platform has just expired. They will be in limbo for a bit. The possible saving grace is 1) they should be able to reuse the hardware, and 2) the new database platform hopefully is more stable than the current one, but it will be a Version 1 product and will probably behave like one as well. Putting a little DATAllegro lipstick on the SQL Server pig does not seem that attractive to me, but time will tell.

    I speak from first hand experience with the DATAllegro product. The company I am involved with had decided to bring it in, but after a dreadful migration to it, we gave up and returned it. Stability was horrible. Functionality was very limiting (great for simple scans, horrible for analytic SQL). Dozens of bug filed. Software upgrades were painful. Needless to say we still have our “legacy/OLTP” platform and have since made some significant improvements to it (amazing how far a little engineering goes). Guess sometimes you don’t know how good you have it until you try something else. Since then I have been made aware that others have also returned their DATAllegro systems or come to a hefty compromise in the customers’ favor.

    Now I am not here to bad mouth DATAllegro as I have nothing to gain by doing so, but do I think there is quite a bit that is not being told. In all fairness Curt, have you anything that is not positive to say about DATAllegro? Any Ws at all in the SWOT analysis?

  5. Mark Callaghan on July 26th, 2008 12:32 pm

    This poses a risk to Teradata and Netezza rather than Oracle. This solution is great at aggregation on single table scans with 100TB data sets. It poses a risk to Oracle when it can run complex queries with many joins over 10TB data sets. Teradata hasn’t taken this market from Oracle, why should we expect this to?

    It should be great for Microsoft deployments because the same underlying server (SQL server) can run the OLTP servers and on the DW nodes, so problems with different data types won’t occur and the DBA tools should remain familiar.

  6. Curt Monash on July 26th, 2008 2:30 pm

    @DW Consultant,

    I’ve said plenty of unfavorable things about DATAllegro right here in this blog. You shouldn’t have far to look for them.

    @Mark,

    So far as I can tell, data warehouses are getting larger somewhat faster than Moore’s Law + software improvements are raising the upper end for Oracle and SQL Server. Thus, the universe of customers whose needs can’t be addressed by Oracle or plain Microsoft SQL Server is likely to grow, not shrink.

    CAM

  7. Stuart Frost on July 28th, 2008 4:30 pm

    DW,

    Thanks for the ‘compliment’. “Ice cubes to Eskimos” – not the most original and certainly not the most insulting thing that’s been said about me online.

    I don’t usually respond to anonymous posts, but thought I’d make an exception in this case because you are really doing a disservice to the great people here. For someone who’s not here to bad mouth DATAllegro, you sure don’t have much that’s positive to say!

    Despite your claims to the contrary, we actually only had one system returned since the company started. Like most startups, we went through a period where the product suffered from some early stage growing pains. This was mostly due to HW reliability when we were building our own servers. We fixed that by going to standard servers and storage, which gave our engineers more time to focus on software improvements and overall quality. Most of our customers stuck with us and are now having great success on the v3 platform. Shame you guys didn’t, but I’m glad to hear your project was a success in the end.

    So, whatever you may think of me (and I’m not sure what I did to incur your wrath), the fact is that DATAllegro has some of the largest, most successful DW installations in the world running on its technology and that’s something the team here is justly proud of. I’m certainly very proud of what they and our customers have achieved together. I’m also very pleased that Microsoft (after extensive due diligence :) ) has recognized our achievements and wants us to help take SQL Server to the next level in the DW space.

    Finally, it’s funny that you think Curt has “fallen for my game” (whatever that is). He’s consistently been the most insightful analyst I’ve spoken to and has often given me a hard time on both technical and business topics. There’s plenty of evidence of this in his various blogs, if you care to read them.

    Stuart Frost
    CEO, DATAllegro

  8. Bill Walters on August 6th, 2008 11:02 pm

    FACT: DATAllegro spent $62 million in venture capital money and has (2) paying production customers. Ice cubes or no-ice cubes this is pretty much all you need to know………

  9. Curt Monash on August 7th, 2008 2:45 am

    Bill,

    Where did you get those figures?

    CAM

  10. Bill Walters on August 7th, 2008 8:19 pm

    Curt,

    The amount of funding received by DATAllegro is public knowledge, $62 Million over 5 years

    http://www.intelligententerprise.com/blog/archives/2008/07/what_the_micros.html

    The customer count comes from interviews with current and former sales people. Notice the qualifiers “paying, and production” $62 million can buy a few customers that do not have budgeted projects. They also had the habit of listing proofs of concepts as customers. But at the end of the day there are a total of two (2) companies that paid for the solution with a license agreement and are currently using the solution in production…….. take it to the bank

  11. Curt Monash on August 7th, 2008 11:03 pm

    Ah. I’d forgotten that blog post, and I hadn’t done the arithmetic myself. Thanks.

    As for the two paying customers — since you’re so knowledgeable, would you care to name them?

    Thanks,

    CAM

  12. L.J. on September 17th, 2008 2:41 am

    Looks like Microsoft/DATAllegro just gave Oracle and the others nearly 2 years (first half of 2010) and they have ceased selling DATAllegro to new customers.

    http://news.cnet.com/8301-10805_3-10042604-75.html

  13. Microsoft/DATAllegro time frame announced | DBMS2 -- DataBase Management System Services on September 17th, 2008 3:14 am

    [...] of a timetable for DATAllegro/SQL Server integration.  Per Ina Fried — with a hat tip to anonymous commenter L.J. — Microsoft says: The final version of that product is slated for the first half of 2010, [...]

  14. Garth Noakes on August 19th, 2009 9:28 pm

    As a long-term DBA, ex-SQL server DBA, Oracle DW specialist (and recently a Teradata Certified Master)

    a) almost all of the article is just hot air
    b) hardware is hardware – fixed limitations
    c) once it is exploited properly, software functionality is the differentiator (but most companies don’t know how to exploit hence the Teradata etc advantage)
    d) ask me if you want an unbiased and professional opinion (but not if you are just trying to sell to your manager)

  15. Curt Monash on August 19th, 2009 9:51 pm

    In the event, Exadata was announced two months after I posted his. Oracle DID have a major internal project going on.

    On the other hand, it’s now almost a year later, and Exadata doesn’t seem to have set the world on fire …

Leave a Reply




Feed: DBMS (database management system), DW (data warehousing), BI (business intelligence), and analytics technology Subscribe to the Monash Research feed via RSS or email:

Login

Search our blogs and white papers

Monash Research blogs

User consulting

Building a short list? Refining your strategic plan? We can help.

Vendor advisory

We tell vendors what's happening -- and, more important, what they should do about it.

Monash Research highlights

Learn about white papers, webcasts, and blog highlights, by RSS or email.