Oracle's CTO: No Way a 'Normal' Person Would Move To AWS (zdnet.com)

Amazon may have turned off its Oracle data warehouse in favor of Amazon Web Services database technology, but no one else in their right mind would, Oracle's outspoken co-founder and CTO Larry Ellison says. From a report: "We have a huge technology leadership in database over Amazon," Ellison said on a conference call following the release of Oracle's second quarter financial results. "In terms of technology, there is no way that... any normal person would move from an Oracle database to an Amazon database." During last month's AWS re:Invent conference, AWS CTO Werner Vogels gave an in-the-weeds talk explaining why Amazon turned off its Oracle data warehouse. In a clear jab at Oracle, Vogels wrote off the "90's technology" behind most relational databases. Cloud native databases, he said, are the basis of innovation.

The remarks may have gotten under Ellison's skin. Moving from Oracle databases to AWS "is just incredibly expensive and complicated," he said Monday. "And you've got to be willing to give up tons of reliability, tons of security, tons of performance... Nobody, save maybe Jeff Bezos, gave the command, 'I want to get off the Oracle database." Ellison said that Oracle will not only hold onto its 50 percent relational database market share but will expand it, thanks to the combination of Oracle's new Generation 2 Cloud infrastructure and its autonomoius database technology. "You will see rapid migration of Oracle from on-premise to the Oracle public cloud," he said. "Nobody else is going to go through that forced march to go on to the Amazon database."

Comments

Wow is Larry ever tired of being wrong?

[Anonymous Coward]

> Nobody, save maybe Jeff Bezos, gave the command, 'I want to get off the Oracle database."

I've never heard anybody use Oracle who wasn't saying that. Every oracle customer I've dealt with has "getting rid of this fucking goddamn shit" as a #1 priority.

Re: Wow is Larry ever tired of being wrong?

[Anonymous Coward]

I heard a story where a company was audited for license compliance. Some very minor issue was found (one thing misinstalled on one client PC, that was not being used). Oracle wanted the company to commit long term to their cloud platform or stop using Oracle all together in 30 days. They did not know this client already had a project to migrate off Oracle that was basically ready (they would cancel Oracle within 6-12 months). They went with it and took the 30 days option, putting extra effort in finishing the last bits. The face the saleswoman made was awesome. Turned out well also, migration was a success.

Re:Wow is Larry ever tired of being wrong?

[Major Blud]

Ditto. It sounds like Larry is taking the same line as his Oracle sales staff, as in if the potential customer doesn't like what they have to offer, insult them and threaten them into buying it or renewing it. I've actually heard them tell people that their "career would go nowhere unless they purchased Oracle".

Re:Wow is Larry ever tired of being wrong?

[ahodgson]

Only if you're running a Fortune 500 and need a multi-server r/w database cluster or a giant data warehouse. For the vast majority of sites PostgreSQL is easier to use and just works.

Re: Wow is Larry ever tired of being wrong?

[UnknowingFool]

Depends on the product and aspect. Oracle DB is fine for what it does. The licensing is onerous and is what causes most customers to revolt. Their other products like ERP can be down right shitty.

Re:Wow is Larry ever tired of being wrong?

[Anne Thwacks]

And you probably spend countless man hours trying to get Postgresql to perform the same way Oracle does.

Or a much smaller number of hours getting Postgresql to do the job you really wanted, rather than the piss-poor job that Oracle was doing.

Costless man hours and wasted resources when you could of paid someone else to do it for you and actually concentrated on selling the product you are producing.
Of course, you still have the option of paying someone else with Postgresql. You just would not have to pay them so much, or for so long, or pledge your firstborn son, or sign contracts with Lucifer.

If you need a database, then dealing with Oracle is the option from hell. Yes, I have managed successful migrations from Oracle to Postgresql. No I wont touch your Oracle installation with a 10 foot pole.

Re:Wow is Larry ever tired of being wrong?

[Anonymous Coward]

For comparison, we have one of our major in-House software products running on oracle (our CRM/ERP) and one running on SQL Server (our POS).

- Oracle "Just works" for years without any problems, while SQL server constantly needs babysitting.
- When an error occurs, the SQL Server error messages are pretty worthless, while the Oracle error message pinpoint the exact cause and the exact location of the error.
- We get maybe 2-3 "unexplainable" problems a week with SQL Server, and maybe 2-3 a year with Oracle.
- For software development things like the "readers block writers" approach of SQL Server gives you a ton of headaches that just "work out of the box" with Oracles locking model of only writers blocking writers.
- A lot of things that require different tools and different languages in SQL server are just built into the standard Oracle SQL engine. For example you can Access Oracle OLAP with standard SQL queries, but you need a different tool chain to access the Analysis Services of SQL Server.

Having said that, I'm also having the "how to possible get away from Oracle" in the back of my head, but that is solely based in the way their licensing terms keep getting worse and worse all the time.

Re: Wow is Larry ever tired of being wrong?

[petergriffinismyhero]

This happened at my work. It was either commit to a seven figure fine or commit to a multi year Oracle Cloud contract we never ever ever would use for 1/3 the cost. Any consideration given to our being a 20 year Oracle customer or that the infraction was ambiguously interpreted? No, because Oracles license Nazi's are just that: Nazis. Of course we are an Oracle Cloud customer now (and have never even logged into the cloud portal), and just as soon as we can get off this POS company's platform in a few months, we will never have to deal with them again. Fuck you Larry.

Re:Wow is Larry ever tired of being wrong?

[mdhoover]

Every oracle customer I've dealt with has "getting rid of this fucking goddamn shit" as a #1 priority.

We aren't getting rid of Oracle DB because of the product (it is solid, reliable and consistent), we are dumping it because dealing with Oracle the company is a f*cking nightmare and they treat you like shit.

Lying like an oracle sales rep proper

[Anonymous Coward]

I've personally architected and implemented a move from two large exadata boxes (abut 1PB, 120GB per day EDW) to a mix of aurora, redshift and gcp's bigquery. It is indeed possible and we were not alone. Just join any AWS ReInvent event and talk with the people you meet there.

The thing that makes moving difficult is Exadata made it extremely easy to write well-performing bad sql, usually powering some OLAP-based BI. Forklifting that crap is not an option.

Thing is, you don't just get databases in the cloud, you get managed ETL, efficient queues, cloud functions, you get well thought IAM (at least in AWS, GCP's is still-but-not-for-long lagging behind), and all of that allows you to rearchitect significantly. We got rid of, for instance, Oracle OBIEE which generates hideously inefficient SQL queries, and replaced it with a mix of google data studio (yeah, that basic) and microstrategy for the analysts that need it.

The migration cost us around 3m eur, and paid for itself the very next year. We had zero infra-related incidents and performance is well above what Oracle offered, cost is about 10x less, and we havent even begun optimizing it.

Last but not least, It was actually pleasant to work with and we had near-zero regrettable attrition among developers during the project. I'd never ever consider working in an Oracle shop ever again, for anything less than enough-to-retire-in-two-years kind of money.

Two other thing to note. AWS has very good support, none of that 'it works as designed, ticket closed' shit. You get greybeards responding to your tickets directly. GCP has somewhat good support but they Really want the enterprise market so once you cut through google's internal bureaucracy and get their attention - it is a breeze. The only notable exception is Amazon. We found that a lot of what's in the documentation is not fully accurate, and scalability beyond proof-of-concept sized applications is nearly always a problem, and some of the problems are wicked. We have since decided to not do any Azure and rely purely on GCP and AWS.

I am a CTO of a 25bn company. I've previously spent 10 years as owner of Oracle-based BI team at a 100bn company with money to burn. I would not exactly call myself a not-normal-person :-)

Re: Not dealing with Oracle = big win

[Major Blud]

Could be. I've seen both scale to thousands of users on equivalent hardware.

One edge Oracle has is RAC. MS SQL has AlwaysOn Clusters, but that doesn't offer the same type of N+1 solution as RAC (not to mention that you have to code around it for it to really be effective).

Thanks for not starting a flame war :-)

Re:Meh.... Two giants bickering

[garcia]

The last time I worked within an Oracle-based warehouse was 2012. Since then, I've been exposed to any number of others, including taking over a on-prem SQL Server warehouse and moving to BigQuery and, currently, deciding how to handle the Redshift warehouse provided to us by a DBaaS vendor.

BigQuery is Petabyte scale, no infrastructure to manage, lightning fast, incredibly inexpensive compared to on-prem SQL Server, and is supported by a ton of toolsets. Redshift is basically the same, with the added negative bonus of having to support it with instances.

While 6 years is an eternity in the analytics space, we're talking about hours-long queries being reduced to single seconds. I'd love to see Oracle be able to keep up with these cloud-DB technologies.

Re:Meh.... Two giants bickering

[HornWumpus]

So, yeah, if you're trying to join the same table to itself 7 times

I worked with that dude! He wrote his master's thesis on SQL query optimization. Didn't know what a query plan was. That's not 'competent' anywhere.

BTW 'You're holding it wrong' is the wrong answer for data warehouses. Relational databases are OLD tech, nobody wants everything different. Many coders are familiar with the options. Keeping copies or summaries of data in memory isn't new. They did that with CICS and COBOL, 'middleware', pinned views, stored procedures, OLAP cubes. So they're calling it 'windowing functions' now? Nice. What's the new word for index? table?