Well, I’m about to start development of a social application (Squirrel2 project) and I have to decide on what technology stack I’m going to use. This is a very important decision: making a wrong technology choice can easily result in project failure.
Here is a good research on the current technology stack of well-known social applications. Looks a bit intimidating, right?
Do I really need all this software? I don’t think so.
All new social applications face two major risk at the beginning:
- Their development may be not complete at all. Most new social applications (with some exceptions like Google+) are developed either by individuals or small teams with very modest financing. In both cases resources are very limited and there may be just not enough time or money to complete development. This is very true in case of Squirrel2 – I have just around 500 hours for the whole development accordingly to my one year challenge.
- Even if the initial development has been completed, the application may not get enough traction with its potential users. This happens more often than not. Therefore developers should bring a first version to production as soon as possible so that to verify its viability before too much money or effort is spent. This implies minimum of the initial development as well
So the bottom line is: the initial technology stack should be such so that to allow launching the first version of the application as soon as possible and with as less investment and development as possible.
Notably, initial versions of many successful social applications were built using technologies that are more known for rather rapid development then scalability or high performance:
- The initial version of Twitter was built in Ruby on Rails.
- The initial version of Facebook was built on PHP.
And what about scalability, ability to serve millions of concurrent users? I wish I had that problem Those issues will be dealt with when and if it becomes necessary.
Squirrel2 technology stack must satisfy following requirements:
- Main development environment/main technology should be optimized for quick development
- At the same time it should have a clear transition path to more scalable solution
- Technology stack must minimize amount of new development; the more ready made solutions I can use instead of doing my own development, the better. I will make exception for technologies that I’m specifically interested to learn, though
- Technologies that I already know are preferable whenever feasible; I don’t have much time for learning given that the whole Squirrel2 development budget is 500 hours
- Naturally, the technology stack should be suitable for development of web applications
I have been thinking of the technology stack for some time. Here is what I decided to use.
Since my main competence as a software developer is in Java, I could choose that language as the main technology. However it would be somewhat boring… I want to learn new things, remember?
So after some research I have decided to use Grails.
A primary reason for this is that it is a rapid development technology. Yeah, I know it was inspired by Ruby on Rails, so should I have used RoR?
RoR is very good technology. The reasons why I decided on Grails and not on RoR are following:
- I just happen to like Groovy (the language used by Grails). I think it yields in a very elegant and easy-to-understand code which I consider very important qualities. Absence of those qualities is the reason why e.g. I did not select Scala and Lift which I evaluated as well; to me Scala code is just somewhat cryptic and difficult to read.
- Groovy produces quite compact code, about twice shorter than corresponding Java code. This is also a very important advantage in my opinion.
- Groovy++ and Java (ability to mix Groovy and Java, to be precise) offer a clean and easy gradual path to future performance improvements in case I would need improve it. RoR does not offer a similar path as far as I know.
- Ability to mix Groovy and Java allows leveraging a huge collection of high quality open source software that Java accumulated over time (remember, my goal is to minimize amount of development).
Grails comes with a lot of plugins that provide functionalities I hope to leverage. Grails community is vibrant. In short, Grails is “what the doctor prescribed” Decided: my choice is Grails.
Grails is mostly used with relational databases. I could use e.g. MySQL which have worked well for me in my previous projects. However this time I have something else in mind – namely, NoSQL DBs.
I think NoSQL DBs are a part of a major paradigm shift happening right now in IT industry and I want to have hands on experience with those technologies. Funny enough, unlike much hyped Web 2.0 or SOA this paradigm shift, which in my opinion is much bigger and more important than those two, happens almost unnoticed. I will write a post or two on the paradigm shift and NoSQL DBs soon.
Hbase is for huge data grinding facilities; it is too big for my needs. Gemfire is proprietary, so I am left with MongoDB, Redis and Riak each of which is quite good. I will think a bit more of this and I will describe my choice later in a separate post.
Naturally, social applications use web technologies at their presentation layer. So will be Squirrel2 doing. I plan to use HTML5 and Rich Internet Application technologies.
That’s all for today. I will describe my choice of tools and supporting technologies in the next post.