Slashdot Mirror
Amazon Plans To Move Completely Off Oracle Software By Early 2020 (cnbc.com)
Amazon plans to be completely off Oracle's proprietary database software by the first quarter of 2020, reports CNBC. The plans come after the company moved most of its infrastructure internally to Amazon Web Services. From the report: Amazon began moving off Oracle about four or five years ago, said one of the people, who asked not to be named because the project is confidential. Some parts of Amazon's core shopping business still rely on Oracle, the person said, and the full migration should wrap up in about 14 to 20 months. Another person said that Amazon had been considering a departure from Oracle for years before the transition began but decided at the time that it would require too much engineering work with perhaps too little payoff. The primary issue Amazon has faced on Oracle is the inability for the database technology to scale to meet Amazon's performance needs, a person familiar with the matter said. Another person, who said the move could be completed by mid-2019, added that there hasn't been any development of new technology relying on Oracle databases for quite a while.
Oracle can keep circling the drain.
Ellison was initially a big critic of cloud computing and famously boasted that it would be a flop. Now that a lot of companies have embraced it, Oracle is left scrambling. Workday, Amazon, Google, Microsoft...they all have a huge head start.
If this keeps up poor old Larry is going to have to sell off one of his Hawaiian islands or a couple of Malibu estates...oh the humanity!
I didn't think Amazon would be using the evil corp products aka Oracle.
Most companies get sucked into the maelstrom by adopting an ERP system that requires Oracle DB as a backend. Then they figure "Hey, we have this DB, so we can use it for everything!" Big mistake.
Whenever I meet someone that uses Oracle ERP, I always ask them if they are happy with it, and would they still make the same decision if they could do it over again. So far, this many people have said yes: 0.
> The primary issue Amazon has faced on Oracle is the inability for the database technology to scale to meet Amazon's performance needs, a person familiar with the matter said.
This is how big tech companies do a BURN!
And it couldn't have happened to a nicer megacorp.
Whenever I meet someone that uses Oracle ERP, I always ask them if they are happy with it, and would they still make the same decision if they could do it over again. So far, this many people have said yes: 0.
Nobody is happy with their ERP.
Oracle screws their customers and pulls absolutely nonsensical licensing demands. Example: You have 1 tiny VM with 1 virtual processor runnning oracle... but the VM runs in a cluster with 1000 cores that other VMs are using. Oracle will demand you license 1000 processors of Oracle for that 1 VM. It's the most insane logic you've ever heard in a licensing discussion.
The only good Oracle is the Oracle you don't use.
You clearly never asked the DBAs and Sysadmin's who have many years of long and frustrating work bashing their heads against Oracle if they're happy with it. They'll tell you no, but then if you ever get the truth out, "I've got a job for life, this shit is so fragile, they can't fire us."
How do I know? I was a BOFH dealing with about a dozen servers for a mid sized company's Oracle ERP system. They were just RHEL boxes, nothing too special. Generally shit just worked. Stupid simple shit for a Unix sysadmin. I pretty much coasted along for a decade at that job, showed up four hours a day, three days a week. Getting paid $150k for it at that. I pretty much did fuck all at that job. Only reason why I'm not at that job anymore? Company went bankrupt, otherwise I would still be slacking at that place.
So, yes, I was happy with that ERP system, just not for the reason people think you'd be happy with it.
"Oracle still calling Amazon four to six times a day to sell services"
"Amazon has blocked 38,000 individual phone numbers in attempt to avoid Oracle sales calls"
"1 in 4 Amazon employees job description includes 'Telling Oracle to piss off' to help deal with never ending sales calls"
"Oracle buys AT&T in order to get cheaper rates when calling Amazon"
"Oracle ordered by federal judge to stop stalking Amazon"
and so on...
They're moving to non-relational databases, such as S3 and DynamoDB. The problem with relational databases is that they can't scale beyond a single host. This follows from the CAP theorem: Unless you are willing to sacrifice some amount of consistency or availability, you can't have partitions. S3 and DynamoDB support limitless horizontal scaling because they use eventual consistency.
The trade-off with going down the NoSQL route is that you no longer have the concept of transactions, and you have to write your software in a way that will tolerate tables being in an inconsistent state. However, the advantage of this approach is that your service will always scale. Therefore, at Amazon, they encourage you to always use NoSQL, because if you choose a relational database, you're assuming that your software won't have to scale.
Microsoft, Oracle, etc. They all are the corporate equivalent of a cult, very similar to the big abrahamic revelation cults ("religions"). "Here, have some flaky lock-in software. It comes in shiny boxes and with flashy names on it. And I'm wearing a suit and it's really expensive and complicated, so it's very very professional."
You get miniature versions of this in the web world as well. I'm currently maintaining a mid sized brand website that is an utterly unbelievable Hodge-podge of commercial WordPress plugins. A true nightmare. But even thinking about doing the same with some totally fucked up Oracle installment just about creeps me out even further. ... At least I'm dealing with FOSS and can implement my own models without having to buy some extra schema contingent or something.
We suffer more in our imagination than in reality. - Seneca
...imagine the chances of a normal sized company to do so.
We moved all of our Oracle instances to PostgreSQL. The biggest problem used to be ESRI, but their PostgreSQL support is now really good. And as a bonus, performance has gone up.
All of our in-house software was migrated to PostgreSQL reasonably easily. Of course, I saw the proprietary database trap back in the 90's, and went with PostgreSQL from the start. I didn't so much talk management into using it as I just used it and didn't bother explaining it to anyone unless they asked.
S3 is object storage, not a database. While AWS offers front-ends like Athena to query it, it's still a file system. You can persist lots of data pretty economically, but it's not a database. You do get benefits like automatic replication across regions, archive policies, etc. If you consider S3 a database, I guess you would have to consider NFS one as well.
DynamoDB is indeed a NoSQL database, and, yeah, I would not use it for transactional database operations. Amazon hardly always encourages NoSQL, though, they offer RDS (hosted MySQL, PostgreSQL, MSSQL and even Oracle) as well as their own relational solution Aurora, which is a highly scalable MySQL and PostgreSQL compatible database and supports transactions.
If Amazon encourages anything, it's to leverage a combo of RDS/Aurora for transactional work, DynamoDb or Elasticache (Redis) for key-based persistence and then Redshift and EMR (Hadoop/Pig/etc.) for warehousing/lakes/analytics; and then using leveraging things like Lambda and SQS to "glue" things together. Of course, leveraging that entire stack can easily lock one into their ecosystem (which Bezos won't shed any tears over); but along with the other gazillion service offerings they offer, it is pretty comprehensive.