4 ms·
I tried using simpleDB for a while, and would not recommend it for production use. No joins - SimpleDB was designed for parallel operations - so if you need
by ptm 19y ago
I tried using simpleDB for a while, and would not recommend it for production use.
No joins -
SimpleDB was designed for parallel operations - so if you need data from multiple tables, you need to fire multiple queries (in parallel preferably).
No retrieval of multiple attributes -
You cannot retrieve attributes from multiple items. So if you need to retrieve 10 items, you'll have to first fire the condition query to get the item-ids and then 10 queries for each item. If you're using serial access, this could be painful.
UTF-8 issues with the REST interface -
Minor annoyance. It works with the SOAP interface though.
1024 byte limit for each attribute -
This could sometimes be an problem, especially if you're using UTF-8 data (with the REST interface issue)
If you're ready to code for parallel access, no UTF-8 (while using REST), and don't mind the 1024 byte limit then you could give it a try.
- sadiq 19y agoI wonder if there's potential for offering some kind of SQL engine that sits atop SimpleDB. In the same way that MySQL used to sit atop of BerkeleyDB. Could even run the engine on EC2.
- xirium 19y agoFrom user ptm's description ( http://news.ycombinator.com/item?id=133086 http://news.ycombinator.com/item?id=133086 ), you could probably hack up the MySQL NDB Cluster interface.
- andyjenn 19y ago...or a MySQL engine on top of S3? http://fallenpegasus.com/code/mysql-awss3/ http://fallenpegasus.com/code/mysql-awss3/
- sadiq 19y agoI was thinking of an actual business idea. Providing some kind of higher level abstraction in a pay-as-you-go utilisation, hosted on ec2 and maybe backed on to simpledb. You'd really want to throw memcache in there as well. I wonder how easy porting that to memcache and simpledb would be?
- jdavid 19y agothe problem with simple ideas for aws, is that you need to either pay via data, bandwidth or instance hours. i would love to pay for process hours on some software. i wish there was standard way for aws customers to exchange services based on process time. in all likelihood amazon has done some of this work for simpleDB, but it would be great to build a piece of code on one linux box and then to sell that time to other users. i could think of a number of situations where I would love to pay for process time instead of instance time. email, databases, caches, load balancing, etc...
- jdavid 19y agothis looked interesting too http://thrudb.org/ http://thrudb.org/