Amazon's Consumer Business Has Turned Off Its Oracle Data Warehouse (bloomberg.com)

An anonymous reader quotes a report from Bloomberg: Amazon.com Inc. has taken another step toward eliminating software from Oracle Corp. that has long helped the e-commerce giant run its retail business. An executive with Amazon's cloud-computing unit hit back at Oracle Executive Chairman Larry Ellison, who ridiculed the internet giant as recently as last month for relying on Oracle databases to track transactions and store information, even though Amazon sells competing software, including Redshift, Aurora and DynamoDB. Amazon's effort to end its use of Oracle's products has made new progress, Andy Jassy, the chief executive officer of Amazon Web Services, tweeted Friday. "In latest episode of 'uh huh, keep talkin' Larry,' Amazon's Consumer business turned off its Oracle data warehouse Nov. 1 and moved to Redshift," Jassy wrote. By the end of 2018, Amazon will stop using 88 percent of its Oracle databases, including 97 percent of its mission-critical databases, he added.

Comments

Hurrah

[TimMD909]

Just when I was getting unhappy about the refurb Mac situation, Amazon brings me back by saying "fuck you" to Oracle. I honestly don't know what to feel now...

Re:Hurrah

[Anonymous Coward]

Well, in Amazon's defense, every single business on earth wants to attain a monopoly and then raise prices while cutting quality. That is THE prize on which they all have their sights set.

There are no exceptions.

Re:Hurrah

[Anonymous Coward]

You seem to draw a line between good and bad and think that everyone on the good side likes each other and everyone on the bad side likes each other. It doesn't work that way. The fact that Amazon and Oracle both suck doesn't imply that Oracle must be good for Amazon. They suck for Amazon too.

When you stop drawing the mentioned line you can learn to accept that a friend of a friend isn't necessarily your friend, a friend's enemy isn't necessarily your enemy, and an enemy's enemy isn't necessarly your friend. And also that a bad person or company probably isn't bad in every respect, just like a good person or company probably isn't good in every respect. A terrorist, once caught, may be perfectly honest about his motives because he doesn't lie out of principle. A bad company like Amazon may have the same issues with another bad company like Oracle as anyone else has. Reality isn't binary, it's complex.

Re: Hurrah

[c6gunner]

True. I wrote a couple reviews warning consumers against items which were obviously fraudulent; Amazon's response was to delete my reviews and send me an emailing warning that "repeated abuse will result in you no longer being able to post reviews". Apparently telling people that a "4k" camera isn't really 4k, and that a "2,000 watt power converter" won't handle more than 200 watts are both considered "abusive".

Re: Hurrah

[Anonymous Coward]

And many businesses, including banks. And lots of other places beyond that.

Your main choices are basically Oracle, MySQL (Oracle), Microsoft (and Windows costs), IBM, or open source and working out some sort of support:risk tradeoff, if you want something directly relational. It puts Oracle in a pretty good position.

Re:Hurrah

[gmack]

Legacy applications. It wasn't that long ago when MySQL didn't handle load very well at all, IBM's enterprise database and Ingress (doesn't really exist now) weren't being marketed at all, and the other competition didn't even exist. Many customers are also sold on the idea that Oracle DB runs best on Oracle's OS, running on Oracle's virtualization system (actually virtualization on anything else can can cause licensing issues regarding Hyper-threading) running on Oracle's hardware.

It is not easy to switch away from Oracle, I man sure it's all SQL, but all of the databases have different quirks and extensions so it can take many months of conversion and years of testing to know if the conversion went well. Now imagine you are a business with some critical app. When things are good, there is no point in spending all of that money to make the conversion, and when things aren't, well it's just faster to buy more licenses so the needed software.

And of course, Oracle knows this and don't even pretend to be nice about it. Their sales department is well tuned towards bleeding revenue from existing clients even if it pisses them off or they have to threaten a lawsuit. I worked at a place before that ended up on the incoming end of an audit that bled us for more money. After the audit, the order came down: NO new Oracle projects. We also had issues where Oracle wouldn't sell us upgrades to our Blade system without the purchase of a support contract with penalties for the years the hardware wasn't covered. (dumped it all when I pointed out that the blades were 10x the price of a 1u rack mount server with the same specs) After both incidents, Oracle sales were shocked that we weren't going to expand any of our Oracle stuff.

The existing customer base all pretty much hates them, and I suspect that if Oracle hadn't gone one step too far and bragged about Amazon, their legacy stuff would have been left running for years longer.

Re:Hurrah

[gmack]

Sometimes when a corporation gets too big, they start to feel entitled to their income and stop trying to earn it via persuasion. Oracle is one such company. They have been sued multiple times for defrauding the US government and have no problem threatening their own customers if it means making more money this quarter.

Re:Hurrah

[jpaine619]

There are no exceptions.

Bullshit.

There are plenty of people out there running a business they own because they love what they do or they love the results of what they do. Your blanket statement is just as stupid as any other blanket statement.

Go back to your basement, asshole. It's obvious you haven't a clue what happens in the real world.

Re: Hurrah

[Anonymous Coward]

If you're talking about the relational database product, MS SQL is way more appealing than Oracle. We have both and are migrating away from Oracle.

MS SQL is a solid RDBMS. Oracle has the edge in certain features that are rarely used, and a a tiny performance edge. But the ease-of-use and licensing cost of MS SQL just blows Oracle's RDBMS out of the water. SSMS is so much more modern than SQL Developer, and gets better all the time.

Re: Hurrah

[reanjr]

For most businesses, the operational cost of SQL Server is dwarfed by Oracle's. Oracle only begins to make sense once you've already started hiring an army of people to manage your data. Most companies just never get there. And those that do can often do what Amazon did, and hire a team of open source DBAs.

Re: Hurrah

[kenai_alpenglow]

Another source of loss of wealth was Grant & Co. They didn't only want to conquer the South, but destroy it. They would burn every city they came to in the Vicksburg/Atlanta campaigns. Some of the stories from that time were pretty despicable. I had the opportunity to talk with a very old lady in Port Gibson (supposedly "too pretty to burn"); her grandmother told her of how they had to hide anything that had any worth due to the pillaging and looting of the union army. So by the time the War Between the States (it was never a civil war) was over, the South had lost a good percentage of its infrastructure. And unlike Europe post WW2, there was no "Marshall Plan" equivalent. Instead we got "reconstruction" (i.e., colonialism).

Fuck Oracle

[bobstreo]

some good products, some not-ready-for-primetime (junk) products, pricing and support suck balls.

I hated dealing with them before they bought Sun. Then I actually started to hate dealing with them more than Microsoft.

Kudo's to Amazon. Hopefully they'll start giving away their DB just to stick it to Oracle some more.

Sounds like a new business for Amazon

[Proudrooster]

If Amazon can actually survive Black Friday and the Christmas shopping season on their new database, they might be able to sell it to others who are trapped by Oracle. It would be interesting to know the back story on how much pain and suffering was required to leave Oracle. My suspicion is that they forked PostGresSQL and Amazon enhanced it. Can anyone comment on the details?

Re:Sounds like a new business for Amazon

[Anonymous Coward]

If you've ever used Redshift, its as if they forked PostGreSQL and removed features.

Re:Sounds like a new business for Amazon

[Anonymous Coward]

Redshift has been a great data warehouse for a long time, an MPP variant of Postgresql (fored from ParAccel) unlike the monolithic Oracle. What locks folks to Oracle is the tech debt and the migration effort. What Andy Jassy is saying is the they finally got rid of the tech debt.

Don't ridicule your customers

[Tailhook]

Pretty basic concept. Self evident to most people. Not Larry, apparently. It's amusing to consider that inside Larry's mind he believes that dishing on Amazon's database products will attract more customers to Oracle.

Re:Don't ridicule your customers

[93 Escort Wagon]

Seriously - what the heck was he thinking? He should’ve just kept his trap shut and continued to collect the big bucks. And his sales team could’ve quietly used “you know, Amazon relies on our database products for its mission-critical systems” as a major selling point.

But no, go ahead and drive them away, Larry...

Re:Don't ridicule your customers

[petermgreen]

How big is big?

postgresql seems to be the go-to FOSS relational database for those who care about data integrity, it's been around for decades and there seem to be plenty of paid support options if you need them. I am told it doesn't scale as well as Oracle or DB2 though.

Amazon's name is worth way more than their fees

[raymorris]

Many people think that for really big databases you need Oracle, for handling lots and lots of transactions. MySQL and Postgres are great for smaller businesses, but our business needs Oracle, they think.

Monday morning I'll be telling my boss that even AMAZON doesn't need Oracle. Facebook uses MySQL / Cassandra. If it can handle Amazon's volume, it can certainly handle ours. That's the big cost to Oracle - the press, the realization among other companies that even at Amazon's size there is no reason to use Oracle.

Oracle would do well to GIVE their products free to Amazon and Facebook just so they can say "we power the world's largest companies and databases".

Re:Amazon's name is worth way more than their fees

[Anonymous Coward]

Not so fast there, hotshot! Just because Amazon can replace Oracle with Redshift doesn't mean you can. Amazon not only has the engineers who creating the thing, they also have an effectively unlimited budget.

That means they can easily fix their own bugs and know how to tune it. If it's going to slow they can always get a bigger computer or build another datacenter.

When you run into a problem with Oracle you have a veritable army of high-priced consultants who have probably seen your problem before and can help you fix it. When you run into a problem with Redshift you can turn to an Internet forum. If you need more performance you can't just build a new datacenter.

When considering whether you can use it to replace Oracle because Amazon did, you have to also take into account the fact that you don't have their expertise or budget, and maybe the reason it works for Amazon is that they can bring to bear resources that you can't.

dom

Re:Amazon's name is worth way more than their fees

[dromgodis]

Monday morning I'll be telling my boss that even AMAZON doesn't need Oracle

Ah, but they do. It says so right in TFS. By the end of 2018 they will keep using 12% of their Oracle databases including 3% of its mission-critical databases.

So Amazon can replace most of its database use, but for mission-critical stuff, they still need Oracle. Score another point for Larry.

Re:Amazon's name is worth way more than their fees

[Voyager529]

Many people think that for really big databases you need Oracle, for handling lots and lots of transactions. MySQL and Postgres are great for smaller businesses, but our business needs Oracle, they think.

That’s only a part of the story. Amazon has an unfathomable amount of resources at their disposal. They can afford to say “we are ditching Oracle”, and write checks and hire developers until those Oracle instances are gone. They don’t need support because they can develop in house until it works.

Most businesses don’t have that. What they do have, are upstream vendors, and what that vendor wants, that vendor gets. Even if Postgres provided true drop-in support, the upstream vendor demands Oracle and getting support is a losing battle on anything else.

Given Amazon’s deep pockets, that’s all but an expectation that they would roll their own. Not everyone can do that.

Re:Amazon's name is worth way more than their fees

[munch117]

Except it's the other way around: They're transitioning the mission-critical stuff first. 97% > 88%.

Which makes sense: The big stuff that's in active development, that's what you move first, because that's where you get the most bang for the buck. If you have one particular program running on thousands of servers that each require an Oracle license, then changing that one program could give you huge licensing savings.

On the other hand, all the rubble, the ad-hoc stuff that only runs on that one server in the Timbuktu office and the source code is lost or unreadable? That one you just leave running for now and start looking for ways to make it redundant.

Re:Amazon's name is worth way more than their fees

[lgw]

I just can't understand this. Amazon is Oracle's customer, you don't speak to your customers like that in any business If I had an interest in Oracle i'd call for him to be fired over it.

The way Oracle usually speaks to it's customers is "nice business you got there, pity if an audit were to happen to it". Then force you to buy some $2M product you don't want or need to avoid an audit that would cost you $3M to comply with. That's the entirety of Oracles cloud business, from the rumors I hear.

Re:Amazon's name is worth way more than their fees

[swilver]

They don't, that 3% is probably legacy crap that is being phased out. No need to migrate it first.

Re:Amazon's name is worth way more than their fees

[HornWumpus]

Oracle coders are just worker bees.

Oracle marketing is pure evil. I've seen 'decision makers' pick Oracle, then take a no show job from Oracle at 10x previous pay (for a few years), then retire. SOP for Larry and co.

Good for some stuff

[Anonymous Coward]

Amazon guys used Oracle for a reason: I've heard them say, basically, that it's great until it's not, but once you reach a certain level of need for scaling your ops time scales with the size of your deployment and that just becomes unsustainable.

That's the mission critical reason for Amazon to move away from Oracle. But they also get two major strategic bonuses: (1) they make competing products and the transition away from Oracle is good for their advertising and bad for Oracle's, and (2) money. Cheaper to build in-house on their own products exactly what they want than to be dependent on Oracle in a way that scales with Amazon's business. Unless Oracle can deliver a better product than Amazon's engineer's long-term, there's no reason Amazon should pay Oracle long-term. And Amazon hires some good engineers.

Oracle's glory days have passed

[Anonymous Coward]

As a 25 year veteran Oracle DB admin, Oracle is fast becoming an anachronism in the modern fast paced world of NoSQL and and serverless cloud tech. I remember when the rage back in the early to mid 90s was to have huge data silos, you bought big tin, built big monolithic databases and you processed things on the biggest hardware you could. Then FOSS and cheap hardware started to appear plus the idea to move business logic from the database and put it back into the apps, thus the DBs started to become bigger and more "stupid" just simple big blackbox data stores.

You don't need huge tin anymore, you don't need huge DBs for new projects, with cheap afforable scalable tech you build proper scalable architectures that can make use of NoSQL or RDBMS tech like PostgreSQL ( Redshift is simply PostgreSQL on monster steroids ). Larry used to be a great visionary, I remember back in the mid 90's Larry said that soon everyone will have terminals connected to huge networks, we won't need PCs in every office and home. People laughed their arses off at him but here we are almost 2020 and we all have tablets and mobile phones connected to the biggst network in history with datastores with the whole knowledge of human understanding at our finger tips. Larry is making Oracle DB a self-managing system and that will cause many like me to move on, if there's little do with maintaining the DB let some £5/hour operations dept out in India look after it, my company needs me to move to more interesting things as they want more value for money from my skills.

Oracle and SQL Server are good systems but they're now simply just another choice and no longer the only choice. I love the new plethora of DB choices, nosql DBs and serverless tech from the cloud providers who also supply the supporting tech like on demand scalable processing engines like Lambda(AWS) and Athena(AWS). RDBMS has it's place, it's good solid, trusted reliable tech but it's simply just another choice. Amazon have seen that you don't need big tin, just lots of small scalable tin and you can process more data in 24 hours than you would in an entire year, store more data than ever before. Times are changing and it feels like the 90's again in IT tech, so much change and so many exciting opportunities available right now, it's why I wanted to work in IT tech and it's great!

Vendor lock-in is mindset as much as technical

[DidgetMaster]

It can be a huge pain to migrate an existing application from one vendor to another. A company I worked for moved from SQL Server to Postgres (because of licensing issues) and it took 3 years even though they are both 'SQL'. But even if you have a long laundry list of items that make it economically attractive to move from system A to system B; you still must fight the politics of 'not invented here' or 'we have always done it this way'.

I have invented a whole new kind of data management system. It handles unstructured data way better than traditional file systems. It does a bunch of NoSQL functions better than other systems. It is about twice as fast as SQL Server, Postgres, or MySQL at relational databases. But I have a terrible time convincing early adopters to give it a serious look. I had a potential customer who was having a big problem with Cassandra. They had a table with a couple hundred million rows and periodically they had to delete about 10 million rows out of it. (Cassandra is apparently built to ingest data but not to update or delete it very well.) The operation was taking them 2 weeks to complete. We put the same data in my system and it completed in just 17 minutes. Yet their management would not even consider adopting this technology and could not even give one valid reason why they wouldn't.