Skip to main content

How relevant it is to consider deployment architecture during the design stage of the application?

During the recent assignment, I encountered this question and I decided to answer through this post.

In the olden days of implementation, deployment architecture was the last hurdle before pushing the application to production mode. I sincerely believe that, the equations have changed now a days. Mainly, because, we are moving away from traditional deployment options like purchasing the hardware and managing the data centers or outsource the data center management to some hosting service provider. 

Today, we have deployment options as commodities. I would probably go one step ahead and say, if you are considering to host your application on platforms like GAE, Force.com you do not need to even worry about usual problem scenarios of scalability, availability and fail-over. Be assured that your hosting platform takes care of these aspects.  But PaaS has its own limitations. However, I am in favor of infrastructure services as a commodity as it gives you the control of your infrastructure. You still design your own infrastructure.

Why I would consider the deployment aspects during my design? Does not it take away the abstraction away?
I do not believe in the paradigm anymore, when I used to claim, write your software once, and run anywhere. Because today, even simple applications require handshakes across many applications/infrastructures and built with many technologies.

The cloud options are the prime business drivers. Organizations already know how and where the ir applications are going to be deployed. Your application design should leverage the advantages of this option.  
For example, When I know my deployment is on AWS cloud, I would choose S3 bucket as my storage option, because, I do not need to worry about the fail-over design separately. More over, there will be certain issues involving around S3 towards my content security. Designing the general solution may not address the problem, I opt to design the solution specific to problem, so that I can keep it simple.

It does not take the abstraction away. The solutions should address the issues, quickly and importantly, in a simple manner. If making the solution more abstractive makes the solution complex, then I would not even try that.

Another example I can cite is, say you need a shared file system and your general application design may consider NFS. In AWS, there are options like Gluster, with which you get shared file system with other deployment advantages that you will have to consider separately and address, if you design the system around NFS.

However,  some deployment aspects like scalability, availability, fail-over are common issues to be addressed, why to keep it for last. You may get better options or better way of dealing with it, if you include it in the application design stage. At least you would get good time to analyze!

Welcome your comments.

Comments

Popular posts from this blog

Key to adopt open source product

Friends, I am working on business solution implementation on open source product called Kaltura. Kaltura is a media management solution and has loads of features that compel any business to take a peek into it. More-over this is the only complete end-to-end open source software available to handle digital assets. But it comes with its own head ache. Considering its open source, its understandable. I feel, handling these would ensure you the success in your open source product implementation. 1. In my opinion, before adopting any open source software, build the capability to deal with the inconsistency bundled in the open source software. 2. I would avoid involving external consultants for 2 reasons.      a. I am not sure, they would bring necessary expertise on to table      b. I fear that there would be little ownership, they will not see big picture of my business (neither I am interested to share it all) 3. Alternative to that is to build the tea...

Secure your application on cloud

Handling sensitive data Define sensitive data for your application. Classify as sensitive data and confidential data. Sensitive data is something like password, credit card account number, something that you should not compromise at all. Confidential data could be your customer’s health record, something that requires your permission before its usage. So, you need to define sensitive data in the context of your application. There are many ways to protect the sensitive data in transit; the easiest way is to use SSL. This is nothing different than handling sensitive data in any traditional application.   However, make sure you apply this rule while designing your application for cloud deployment. Alternatively, you can encrypt the sensitive data and transport. Be noted that any kind of protection you design, will have implications on performance. However this is ignorable considering the nature of sensitive data. If you just want to protect your data from being tampered du...

Object orientation and Service orientation

Object orientation is all about modeling the real world information mainly through encapsulation, abstraction, polymorphism, containment, inheritance and so on. In simple words, an object contains the data structures and methods to change the content state. A typical OO run-time manages the life cycle of objects. You can create as many objects as required as long as the run-time does not over run its memory. Service orientation requires you to encapsulate a process. The operations of a services are interaction points with the particular process. Data structures are the arguments for these operations. Service run-time shall validate the correctness of the process. Service can have multiple invocation instances, not multiple instances of service itself. Example: TrainTicketBookingService involves a process. - Check the train timings //  interaction point - Check the seat availability // interaction point - Issue ticket // interaction point There could be many ticket...