All your database are belong to us: AWS wants every data silo on its platform

AWS has introduced a flurry of new database and ML services at its Re:invent conference, including a migration service targeting every database in an organization, Swami Sivasubramanian, VP of ML (machine learning) gave the data keynote today. He claimed that Aurora, a service that is compatible with either MySQL or PostgreSQL …

[Abominator]

As an application developer who connects to both Postgres DB and Aurora databases, I can confirm Aurora is far slower that a properly setup local Postgres install.

Its dog slow for online transaction processing. To get the best out of it you have to move all your processes close to it. Even then its slower than an in house setup.

Yes, it scales...supposedly, but then I have not really pushed it at scale as its ingestion rate is rather poor.

Also there was a previous puff piece on here claiming in very wormy word terms that Aurora was open source as its Postgres. Well that's a complete lie. The backend is completely proprietary to Amazon.

I don't know where they get their stats. I think they genuinely make them up.

Another team I work with internally is moving off Aurora as they used it from day one for their business and have had no end of pain making it scale. Sort of made 'Postgres' a dirty word for them, but then they have to be reminded they are actually using Aurora.

[Anonymous Coward]

Absolutely

Generally that's true of most heavy DB apps, you need to move as close as possible when you have very high transactions rates or recode to batch things better.

A 25 year DBA the most common mistake I see is people simply think they can move a DB anywhere they like but they never, ever check the work profile characteristics. Often find badly coded apps are processing commits on almost every every transaction by default 'cos the dev got paranoid about losing data. You might get away with it if you're lucky but the DB suffers badly 'cos every commit causes disk block writes to DB logging process, which is usally serial and single threaded! So they move the DB into some cloud or external data centre and next thing the one thing no one expected...network latency kicks on hard with every single little peep between the app and the DB taking 2-3 times longer minimum, the app then grinds to a halt 'cos it's so damned chatty.

PHBs and devs who don't read the real info blinded by PR bullshit about "Cloud is always better....well 'cos it cloud innit!".

Saw this happen only a month ago, people will never learn. Still shouldn't complain, every cock up like this keeps fattening up the pensions of DBAs! ha ha!