I’ve been investigating the Amazon S3 service for while now and even opined about building a scalable application that used that service for data storage instead of a relational database. In my article, Building a high volume app without a RDBMS or domain objects, I talked about using a relatively simple flat file format such as XML to store the data in S3 and combined with the right kind of caching, have a reliable, performant and scalable application.
Well, I’m doing it
I decided to give my idea a real world test run. I’ve gotten and idea for a social network style of application that would fit this role perfectly. Largely a read-only system, a dirt simple data schema, no overwhelming need for transactions and no need to really use session management.
The only real departure I’m taking is that I am going to use dumb domain objects. I started to not use them, and stored my data as XML in S3. I then remembered working on another project where using the JSTL XML tags just didn’t translate well (no pun intended…) into performance. It was just too slow. So I changed directions and decided to use domain objects translated in/out to JSON format.
I could have used JSON directly and dumped the object model altogether. Most modern browsers handle JSON data pretty well, so I could have sent that directly to the browser and be done with it, but I didn’t. Right now the objects aren’t completely dump data objects, which is part of the reason I went the route I did. For the moment I’m going to stay on my chosen path, but it is subject to change.
In my next post I’ll cover the libraries I used, the architecture of the app and how it is coming along.
Don't miss anything, subscribe!
0 Responses to “Putting my non-RDMBS idea to the test”