Performance Aspects of Product Development

Performance aspect of any application plays a critical role for user experience. Slower applications frustrate the customers and customers tend to look for faster application from different vendor.   Its good practice to write non-functional requirements also along with functional requirements. If non-functional requirements are not written as product requirements, it is likely that performance aspects are missed to implement during design and implementation phases of the product/application development.  It usually happens that application performance tuning is carried once the customers using the application start expressing their frustrations. The performance improvement activity is carried out as a separate project itself. Not a good practice anyway. There are several ways to look into products/applications to make it work faster.

Capture performance requirements

This is the first step. Customers inputs can be taken in this regard based on their experience of using similar products in the past. The truth is that the customers will be delighted to give inputs; the only thing is they must be asked for it. I remember many customers had made a list of product improvements but they had never expressed them, when we asked their bad experiences with the product, we were surprised to see performance improvements coming in. Specifically define performance numbers for each of the operations of the product. For ex: File Upload shall be completed within 10 seconds, the report shall be generated within 12 seconds. Deciding on these numbers is tricky here. Few prototypes can be tried to see how fast an operation can be carried out after applying all possible performance improvement guidelines.

Have Network Simulators

Each customer can have different network speed.  Non functional requirements must be written basing a network speed may be considering the worst case scenario.  Now the requirements look better. Ex: File Upload shall be completed within 10 seconds at 256KBPS etc.  It is important that every developer in the team uses the same network simulator for getting accurate performance data.

Design the non functional requirements

The poor performance of the product could be because of its design.  This is simply because there are several methods to implement a given product operation. For example by using different algorithms, program logic etc. A simple understandable example is to avoid multiple reads and writes. Read/Writes are always costlier operations. It could be file read/writes, database read/writes or a hardware device read/writes. So reading once the required information and using the program logic gives better performance over reading the information as and when required. This may not be always true; varies from kind of the product/application being developed. The person who has understood the requirements and more importantly the technology involved can best write the better software design.

Avoid unnecessary program Intelligence

Software has to do exactly what it is supposed to do. No extra things. Developers usually tend to add certain intelligence in the program that end customers never wanted. If you are considering performance requirements as Critical To Quality (CTQs), then not even think of adding a line of code that’s not required to meet the requirements.  For example, if a customer registers himself/herself using a registration web page, you just need to give a confirmation text in the same page. It is not required to send an email and mobile text message unless sending an email and text message to mobile is written as end user requirements.

Measure & Improve

In a performance improving activity, most of the time developers will be spending time measuring the time taken to complete an action. Measurement has to happen following a defined system configuration and defined speed. Some of the important parameters to consider are to note the CPU usage, memory usage etc at the time of taking readings. And all the readings are taken at defined parameters. This will help to get accurate data from multiple systems and more importantly merge the data into a single document.

Conclusion

These are high level guidelines.  Any performance tuning subject falls down to writing faster code. There are several web resources to study how to write a faster code. Involving experienced developers is also another thing towards bringing a good performing product.