During the opening keynote of AWS re:Invent 2014 Amazon's SVP Andy Jassy unveiled a new service, dubbed Amazon Aurora, arguing that traditional database software isn't serving customers' needs in a cloud-centric world.
Aurora is MySQL-compatible and, it is claimed, can providing 5x the performance of MySQL for 1/10 the price of Oracle. There are no up-front costs to start using Aurora, and running it on one of AWS's R3 large instances costs $0.29 per hour on a pay-as-you-go basis.
Is this another 'cloud' hype story, or have Amazon actually identified a niche in the market for their new product?
This discussion has been archived.
No new comments can be posted.
Amazon Unveils their very own Relational Database: Aurora
|
Log In/Create an Account
| Top
| 20 comments
| Search Discussion
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
(Score: 3, Insightful) by Lagg on Thursday November 13 2014, @04:17PM
Have code to show that any of the marketing shit your throw up actually has technical merit Amazon? No? Then go away.
http://lagg.me [lagg.me] 🗿
(Score: 0) by Anonymous Coward on Thursday November 13 2014, @04:25PM
Here's the problem they "solve":
Which surely is a problem, but the solution to proprietary databases surely isn't yet another proprietary database. The obvious solution is to improve the performance of MySQL (or MariaDB, to be independent of Oracle).
(Score: 0) by Anonymous Coward on Friday November 14 2014, @02:48AM
Actually, as all modern programmers know, Amazon releases extensive example code for using all of its APIs. The "code" involved in deciding the "technical merit" of a network service is, of course, the client code that is actually doing something with the service. Have a real question? No?
(Score: 2) by Lagg on Friday November 14 2014, @07:02AM
Please do link me to the extensive example code of the DBMS in question so that I can see exactly how it is optimized for a "cloud centric" world and is obtaining 5x the speed of mysql. I don't care about heavily cached and overabstracted APIs in front of what is otherwise a black box. For all you know this is just them giving mysql a different name and trying to sucker you into it. And given the bizarre quotes and terminology in your post I'm thinking it's working.
I also love the modern programmers bit. Because wanting to see stuff that you work with at a lower level and not get locked into yet another crappy proprietary system is just so prehistoric.
http://lagg.me [lagg.me] 🗿
(Score: 2) by goodie on Thursday November 13 2014, @04:30PM
Amazon controls pretty much the entire stack of app development, except relational databases, until now. Since they have already dipped their fingers into this with DynamoDB, it only makes sense. After all, it's just a service like any other. Heck it can even an integral part of the platform... I can imagine that for most businesses, this may not make much of a difference compared to others rdbms anyway. As long as you can SELECT the data you have an out. But this was also true with Oracle, MSSQL etc.
One question comes to mind though: Can they back up these performance claims? Are there any "shortcuts" taken to achieve them which may not suit all scenarios/businesses? I've seen countless "benchmarks" based on ridiculous test cases and scenarios which made no sense whatsoever, especially in the DBMS and DW worlds. Maybe it'll be a 2x speend improvement and not 5x, it's still something. In any case, the speed arguments are often due to poor application coding rather than DB limitations in my experience... I also awonder if it's based on an in-memory database or a mix etc. Actually releasing these scenarios could be of value for people trying to compare offers with and without the cloud.
It'll be interesting to see what the next move will be. MS already has Azure, but Oracle is still missing out on the SaaS and PaaS as far as I know.
(Score: 0) by Anonymous Coward on Thursday November 13 2014, @06:10PM
Given that AFAIU it runs on their servers, I think it should be trivial to make them true: They just have to allocate more resources to Aurora than to MySQL. Or maybe shape the traffic accordingly if the DB is on a different node than the code accessing it.
(Score: 2, Funny) by Horse With Stripes on Thursday November 13 2014, @11:41PM
As long as you can SELECT the data you have an out.
SELECT * FROM `db_options` WHERE `vendor_id` != 'Amazon' ORDER BY `foss_status` ASC limit 0,10
Showing rows 0 - 9 (10 total, Query took 0.0003 sec)
Users selecting those rows also viewed these rows ...
(Score: 0) by Anonymous Coward on Friday November 14 2014, @02:55AM
Trivially, many DB queries can be translated to simple hash-like lookups, which are likely to be very fast if they use the same techniques that various nosql datastores are using. Presumably they're adding performance by optimizing what is easily optimized, and those are the things that they're measuring in that vapormark. There shouldn't be much trouble speeding up those cases if you're willing to slow down the worst cases slightly, in order to run additional layers of optimization.
Also just by having better automated cluster management they might be able to offer a valuable service to many companies.
(Score: 0) by Anonymous Coward on Thursday November 13 2014, @04:33PM
I guess it depends how you define "cheap". $2500 per computer per year is not exactly cheap.
And since attached article does not even load text without javascript, I'm not quite certain how Oracle was charging $25,000/yr for a GPL database that I can install and use on my Debian box for $0 and maybe $100/yr in electricity. All I need to do is `apt-get install mysql-server` and I get version 5.5 installed.
(Score: 0) by Anonymous Coward on Thursday November 13 2014, @04:52PM
> And since attached article does not even load text without javascript,
The gigaom article works for me with noscript and requestpolicy (and adblock and ghostery, etc)
(Score: 2) by datapharmer on Thursday November 13 2014, @05:34PM
You do know that the oracle purchased product "mysql" and the "oracle RDBMS" database are two different products with totally different performance, right?
(Score: 0) by Anonymous Coward on Thursday November 13 2014, @06:10PM
LOL, he doesn't!
(Score: 2) by tibman on Thursday November 13 2014, @06:51PM
The summary doesn't say RDS. It is worded like mysql costs 10x as much as Aurora. I don't fault the AC for that.
SN won't survive on lurkers alone. Write comments.
(Score: 2) by datapharmer on Thursday November 13 2014, @07:05PM
I understand the confusion, and agree that it is terrible writing. That said in the database world "oracle database" always refers to the oracle RDBMS product. It wouldn't really make sense to say it is 1/10th the price of $0, now would it? Heck Oracle didn't even own mysql until a few years ago when they bought innodb. That aside, shame on the writers for that poor comparison. In other news, I can sell you a vehicle that has 50x the speed of a bicycle at 1/10 the price of a Boeing 747...
(Score: 0) by Anonymous Coward on Friday November 14 2014, @02:59PM
a vehicle that has 50x the speed of a bicycle...
For an average Joe de Vries (i.e., Dutch) that'd be about 20 an hour.
So you have a vehicle for sale that goes about Mach 1?
(hoping for the reply that closes the pedantic tag ;-)
(Score: 2) by c0lo on Thursday November 13 2014, @08:27PM
Licenses for MySQL [mysql.com]. Notes:
So here you have your answer: yes, there are features not included in the GPLed version; yes, they can charge whatever they want (even for the GPLed edition); and yes, there may exist buyers at $10k/yr;
https://www.youtube.com/watch?v=aoFiw2jMy-0 https://soylentnews.org/~MichaelDavidCrawford
(Score: 2, Interesting) by terrab0t on Thursday November 13 2014, @05:29PM
Does MySQL-compatible mean it implements MySQL's SQL implementation bugs [sql-info.de]?
For anyone porting software from MySQL to this, these bugs are required. To anyone writing new software, these are a problem. Has Amazon gone the buggy, truly MySQL-compatible route?
(Score: 2) by tibman on Thursday November 13 2014, @08:30PM
Looks like a lot of those implementation bugs were fixed over five years ago. Some of them are just bizarre. Inserting into a primary key that was supposed to auto increment? Yes, that should throw an error called "Don't contradict your own design." Instead it just happily tries to do whatever you told it which results in undefined behavior.
SN won't survive on lurkers alone. Write comments.
(Score: 2) by aclarke on Thursday November 13 2014, @08:30PM
I'll have to do more reading, but I don't really get this one. As others have pointed out, if companies want to dump Oracle and move to a MySQL-compatible RDBMS, there's MySQL. Amazon provides MySQL RDS instances, which already are reasonably cost-effective depending on usage. For those who want to save money and trade it for the hassle of managing MySQL themselves on AWS, you can run MySQL on an EC2 instance quite easily.
A db.r3.large MySQL RDS instance currently costs $0.365 per hour. That same instance with Aurora costs $0.29 per hour. So yeah, you're saving a bit of money but you're also using a beta release (for now). If it truly is faster, the cost differential will increase as you scale up to large instance sizes so I'll give them that. On that note, the place where this may actually be a great boon is that tipping point where with MySQL one instance isn't quite cutting it and you have to start looking at scaling.
What I personally wish Amazon had done is to work on moving an RDBMS to a more granular pricing model. I just want to pay for storage, CPU cycles, and access, and not have to worry about which instance size I'm using. Even better, I want my data to be distributed geographically across multiple data centres, magically and automatically. And with that, I want a flying unicorn.
(Score: 0) by Anonymous Coward on Friday November 14 2014, @03:03AM
Managed scaling, including automatic crash recovery with no client responsibilities after a failure.
Heavy optimization of best cases, and simple client design guidelines that can be followed to usually hit those cases.
This is pushing more in the direction of your flying unicorn than your MySQL comparison. Give it some time to see how they build the service out based on feedback.