Oracle CEO Larry Ellison Steps Down

mrspoonsi writes Oracle founder Larry Ellison is stepping down as CEO. He will be replaced by two executives. Former Oracle presidents Safra Catz and Mark Hurd will be co-CEOs. Ellison will be the Executive Chairman of Oracle's Board, and the company's CTO. Oracle's shares are off by 3% on the news. "Larry has made it very clear that he wants to keep working full time and focus his energy on product engineering, technology development and strategy," said the Oracle Board's Presiding Director, Dr. Michael Boskin.

Comments

Attention Oracle Employees -

[Anonymous Coward]

For those not familiar with the Hurd game plan, let me tell you how it goes.

1. He will say all the right things in front of employees. Part of the team, all in it together, shared sacrifice.
2. He will make the numbers look good for Wall Street.
3. Your cubicles / office space will shrink.
4. Employees will be asked to work from home
5. Real estate will be targeted as a way to reduce expenses.
6. The stock will look good for a while, then implode.
7. When he's finally run out, the next person in charge will realize the business fundamentals are seriously lacking and will have a major cleanup job to do.

Let the hate fly!

[FlyingGuy]

But here is the problem...

You cannot deny that he built a huge empire on something as banal as a database. Arguably the best RDBMS pretty much ever and where is the competition?

I use Oracle extensively. I am a DBA and of all the alternatives out there ( and I have tried most of them ) the only thing that comes even close is DB2 with postGres running a close third and depending on your POV, catching up fairly quickly. Perhaps postGres will eclipse Oracle one day, but not unless they get some serious money behind the project and that probably won't happen because no one wants to pump the millions of dollars it would take into a project that cannot even fix the TXID problem, and make no mistake about it, it is a problem. Also if someone dumps that kind of money into a project they expect some kind of ROI. There might be a few white knights that have that kind of money but they are few and far between and most can find more worthy causes to spend that kind of money on. Don't get me wrong, postGres is a fine DB but it has some faults that make it not so attractive.

Larry understands how to stitch technology together around a DB better than most anyone else I have seen. Arguably Microsoft gets it, but they are stuck running in the windows universe which despite a lot of progress is still broken. You cannot run MS-SQL Server across hundreds of Intel machines and expect it to hold together, but they ave built and end to end ecosystem and MS-SQL Server is tightly integrated, but you can't drop it on a Z-Series mainframe under either IBM's native OS or Linux. PostGres has the same problem but they are moving to fix that, but I am not sure they really understand the problem. Of the other DB's out there ( Mongo, Hadoop, et all. ) that you can do that with, they don't support things like ACID which, like it or not, is pretty much a requirement in way to many situations.

The facts speak for themselves. If Oracle really sucked as a Database it would not be in the vast number spaces that it occupies. You can cap on Larry Ellison all you want, question his lineage, say he is an ego maniacal asshole, but you have to give the man his due. He built a company that does have the answer to almost all the spaces where a DB matters and he built a business that relentlessly pursues those spaces to the betterment of their stock holders and 98% of the people and organizations that use their products.

Re:Let the hate fly!

[FlyingGuy]

Happy to share, and I have posted a link as well.

So every SQL database, Oracle included, has to have some way of keeping transaction order, which is to say which transaction got there 1st, 2nd, 3rd etc. This is part of ACID and it cannot be ignored. Oracle and others solved the problem by using a synthetic number. Oracle's "number" type is not a recognized IEEE standard like an INT or a FLOAT, DOUBLE etc. It i stored using a proprietary scheme in the DB that Oracle guarantees to be correct when it is accessed and it is completely portable. The last estimate I saw ( I rarely look ) was that pushing Oracle to its absolute limits was that it would take ~ 141 years to wrap around. Personally I am not going to be alive then and I doubt Oracle will be either.

In postGres they use a 32 bit unsigned int to keep tract of this. Now 2^32 is 4,294,967,296 transactions, which is a very large number indeed, but when you get into extremely hi volume transaction environments this can get used up pretty fast, like in a few days fast! Since the number is unsigned and postGres is written in C ( although I don't think it matters ) when you hit INT MAX it wraps back to 1!!! and that is a disaster. In older versions it just kept going and corrupting your data and there was really no way out of it, you were just hosed. In the latest version, the database will only go so far and it will force itself down before it wraps around AND will refuse to come up until the vacuum process is complete. On VERY large tables this can take days!

Now the guys who write PG are no dummies. They recognized this and came up with a process called VACUUM, it does many things and it will reset the TXID and keep you safe. In most every application this is fine. In Extreme transaction environments where you build up billions of rows very quickly you have to set the VACUUM process to it's most aggressive level to keep up with inbound transactions and it just kills performance and your TX rate falls into the basement.

This link explains it better than I can. PostGres Wrap Around Problem