Amazon consumer biz celebrates ridding itself of last Oracle database with tame staff party... and a Big Red piñata Amazon staffers – one dressed as a Big Red piñata – have taken to Twitter to celebrate shutting down their last Oracle database used in the retail side of the organisation. The corporate factions have been in various heated battles over the years – from poking fun at the other in conference speeches to scrapping in the courts …
To be fair Amazon developed the whole lot in house. The "Postgres" moniker is just the version that is API/SQL compatible with any existing Postgres application. They have editions for MySql and DB2 plus interestingly Oracle which they choose not to use when replacing their Oracle DBs.
> Do you believe they developed it without looking at Postgres code - and maybe even reusing some? They don't distribute it, so they're covered....
There would be no license problems even if they distributed something derived from Postgresql code. It is covered by a BSD-style liberal license (https://www.postgresql.org/about/licence/).
It depends how you choose to look at it. AWS are doing to Oracle what Oracle did to IBM and DEC many years ago - providing a more affordable product that meets their customers needs.
Is there any lock-in? In theory there's nothing stopping you moving to a different database if you can convince your in-house development team or external software vendor to support it ;) And while I'm making people laugh, I've just broken two of my Queen records. Oh, how I want to break three
In terms of who is better, IBM << Oracle << AWS << ??? The only constant in IT is change.
...what Oracle did to IBM and DEC many years ago - providing a more affordable product...
Not as regards DEC. We were happily using the RdbVMS database when DEC sold it to Oracle. We went along to a meeting at Oracle, where the explicit message was "Rdb is now going to cost you a lot more money".
Yes - this is in-house Amazon only, within AWS customers will continue to use Oracle if they desire.
There are a few large AWS customers (glances at Salesforce and the US Govt) that might be paying a little more attention...
This is about taking away Larry's right to gloat, taking away Larry's business is still a few years away. And running "Oracle cloud" on AWS a few years after that, assuming Oracle cloud have any remaining customers by then.
There are genuine competitors to Leisure Suit Larry and his Minions in the relational database arena. The biggest problem is not that db X cannot do what Oracle can do is that converting the code is a best tedious process as both dbs almost certainly have non-standard extensions, different tool kits, and variations on vanilla SQL. A conversion that is doable if you are willing to spend serious time, effort, and money to do it.
So the real question is what the Minions do to piss off Amazon?
Maybe they did, but not necessarily.
After all Amazon is now a competitor of Oracle, it makes sense for them take the effort and the expense to cut free of Oracle.
That way they can go to potential customers and thell them "sure you can run Oracle on AWS but, why bother with the expense and the agravation of dealing with Oracle? We have this database that can deal with Amazon's needs without breaking a sweat and you can use it for [I have no idea how much, free?]"
That means that Amazon cloud is suddenly much cheaper proposition.
Also, Amazon may have just trained a batch of consultants ready to help you migrate from OracleDB
I suspec that licence fees for Amazon's infamously low margin shopping business are more than pocket change. They're probably also not knee deep in the some of the stored procedures stuff that keep many companies tied to Oracle. It's also worth noting Enterprise DB has made a business of migrating smaller companies from Oracle to Postgres.
Oracle used to have a significant edge when it came to scale, performance and support but they've basically pissed away much of the technological advantage – for OLAP there are now a lot of alternatives, and pissed off a lot of the customers with the changes to licensing while winding down support.
"Those who really need Oracle DB do not care how much license fee is," you mean all 3 of them. The truth most people who are using Oracle can use a different db even a different type instead of a relational and be fine. The vendor lock-in for most is the sheer PITA of migrating from Oracle to something else not some specific feature.
If Amazon can ditch Leisure Suit Larry there are probably very few Oracle customers who could not move to another db. This is a public relation disaster for Larry as a major, well known customer has publicly ditched them. And this customer is very large. As couple of noted, Amazon can give Oracle customers guidance on ditching Oracle for first hand experience. I would not
"So the real question is what the Minions do to piss off Amazon?"
The general sentiment is covered under a quote from the following article:
https://www.businessinsider.com/amazon-cto-werner-vogels-oracle-databases-2017-4?r=US&IR=T
" While he didn't name names, he chided those legacy companies for what he called "punitive" licensing practices. Those companies make customers predict years in advance for the database capacity they'll need under a long-term sales contract, and may sometimes subject them to intense audits to verify they're not using more than they've paid for.
By Vogels' reckoning, this "nightmare" practice means companies tend to buy as many as 30% more licenses than they actually wind up needing, as a hedge against the dreaded licensing audit, "because it's very hard to predict the future."
When I change companies/contracts, I'm often shocked at some of the database licensing practices that are in-place. Development systems with production licences, systems licensed per core when the box is massively over-specced for its workload, applications licensed per seat when every user needs access or per core when there are only a handful of users. And these same companies often have to buy more licences in the event of an audit because they are doing something not covered by their existing over licensing.
Oracle has been loosing ground for years. Its debatable whether at the very top end they are still at the top, but for most normal and havey relational database usages, they have been overtaken in ease of use by pretty much everything else in the market.
Rather than accept this and try to do something innovative about it, they have doubled down on screwing over their existing customers.
Only an idiot would develop a new product using an Oracle back-end.
but once it has been done it can be done again far more inexpensively and much faster. Therein lies the danger of being as aggressive as Oracle has been with it's pricing and contractual fine print. Oracle build a great product but you ca only push your customers so hard because at some point something else will come along that can replace your product. I'd suggest MS should rethink its desktop mess or one of the behemoths might decide to put MS Office in it sights. LibreOffice has all the required elements and already does all that most people need.
"Experience of one company using it's own programmers to rewrite it's own applications to run with it's own database has zero relevance to anyone else enduring same task."
I accept your argument that Amazon migrating away from Oracle can be applied as a general case. Amazon were reportedly paying US$10m for Oracle licencing (I can't find the source...), so there was a significant financial incentive combined with Oracle effectively doubling Oracle on AWS/Azure licence costs in 2017.
However, it depends on what the database is used for - while many applications do have complex database designs that are difficult to migrate, a lot of in-house systems are often much simpler, having evolved from standalone systems (i.e. Excel/Access) and data warehousing/reporting/analytics requirements often differ from production applications.
From AWS's point of view, the costs of moving to the cloud are not small so saving customers money on the way in increases adoption. AWS don't have to migrate 100% of Oracle DB's away from Oracle to achieve this.
On the other side, Salesforce (reportedly paying US$300m for a 9 year Oracle licensing deal in 2013 or $33m/year) likely have a similar viewpoint to Amazon, where the cost savings could be significant. Or at least sufficient to justify any pain.
decided to shoot itself in the foot and move away from leading database product
nah, I suspect that the real engineers who deal with large distributed systems at Amazon got tired of Oracle's inability to provide so simple data consistency guarantee like actual serializable isolation.
Actually, Amazon's stuff mostly works plus they have decent customer service, and you have never been able to say that about Microsoft.
As opposed to people like NewEgg, who have sent me several wrong items, one of which I've been trying to get refunded since December. Or Dennis Kirk and BikeBandit that have sent me wrong motorcycle parts several times. Or Twisted Throttle that refused to sell me parts to move their accessories from an old bike to a new bike.
I could keep listing "top rated" companies that have fallen short time and again.
Amazon just takes my money and sends me what I ordered. It's no wonder they're steamrolling everyone else.
... Or they could use the money saved to pay some taxes. It's been reported that Alibaba paid some $7.7 Billion in taxes last year:
https://www.asiatimes.com/2019/04/article/alibaba-paid-51-6-billion-yuan-in-tax-last-year/
I wonder how much Amazon paid.
You may be right for some features of the very high end. Otherwise Postgres has most definitely caught up in many areas and has its own advantages in others, not least in the flexibility of the backend. Ever since Sun bought MySQL there was an uptick in companies actively engaging with Postgres either through code, sponsoring feature development or paying for professional support from companies that do develop the DB. And after Oracle bought Sun that uptick really ticked up.
The fact is that many companies buy something like Oracle believing there going to need every feature only to find out that they don't actually need them and that there are cheaper alternatives, even after taking the costs of migration into consideration. Companies like NTT wouldn't be playing around with Postgres if it didn't work for them.
"Companies like NTT wouldn't be playing around with Postgres if it didn't work for them."
Telco companies are often willing to use the nastiest most horrible to run solution if it saves them a few cents across their whole estate. Just for instance Open Stack - whereas very few "normal" companies would take that pain and overhead. For those of us that don't want an army of staff all editing text files, then there are easier solutions.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
