Toby DiPasquale on 15 Dec 2007 01:07:09 -0000


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: [PhillyOnRails] Amazon's SimpleDB, let's talk about it


On Fri, Dec 14, 2007 at 04:47:56PM -0500, Evan Weaver wrote:
> Hi Philly,
> 
> SimpleDB is public now. It seems to fall somewhere between a RDBMS and
> Couch, and it runs only on Amazon Web Services.
> 
> Some links at: http://del.icio.us/popular/simpledb
> 
> What do people think?

It must be cool because its built with Erlang ;)

http://www.satine.org/archives/2007/12/13/amazon-simpledb/

The limitations make it somewhat impractical for use at this time,
particularly the 10GB data limit per domain (re: table). That's a
showstopper for us, as you can't cross domains in a query and our dataset
dwarfs that size easily.

Also, I can definitely seeing it being somewhat expensive, given that you
have to pay for storage at a much higher rate, bandwidth into the service
and "machine time" which for right now is a black box. I wouldn't be super
surprised if it was more economical to just run MySQL on a small instance
(though I'd be surprised if it was less economical than running 2, unless
your dataset doesn't fit in SimpleDB).

The API appears to be a bit anemic in terms of query functionality, as
well. I can definitely see a potential porter of some app looking at this
API doc (which is the smallest of all aside from SQS, btw) and thinking
"shit, this is going to be a lot of work"

I'm pretty happy with it, though. When they first proposed the question of
database functionality on the EC2 forum, some of the responses from the
community worried me a lot. Those responses clearly misunderstood CAP
conjecture and fundamental database realities and I was worried that
Amazon would try to do too much in an effort to please those kinds of
requests. Gladly, they put together an API and a service with a clear
scaling and pricing model, if a bit expensive and limited at this time.

Summary: Its BigTable/HBase/CouchDB with an integrated query language
instead of a generic map/reduce framework. Its also a pretty expensive
pay-as-you-go at this time with a fairly small cap on effective data size.
Still, it looks promising and Amazon has a great history of improving
rapidly on AWS.

-- 
Toby DiPasquale
_______________________________________________
To unsubscribe or change your settings, visit:
http://lists.phillyonrails.org/mailman/listinfo/talk