Performance issue solve it by throwing more hardware?

This is the first reaction for most of the organisation against a performance issue. As hardware becoming cheaper day by day the first reaction was to add more hardware.
There are two issues: First the TCO of adding hardware is much higher then the hardware cost itself, if we include services, space and software licensing cost. Second blindly throwing more hardware does not solve performance issue sometimes it make the matter wrorse. In a classic case a mid tier corporations tried to solve their performance problem by adding couple of more CPU to their enterprise servers. To everybody’s surprise the performance of the application got degraded. The server was connected to an enterprise storage subsystem. As the server get upgraded it send more request to the back end disk subsystem, and queues build up in storage system and the response time of the disk subsystem increased and as a result the entire applications response time got increased. A Virtual infrastructure model could have saved the corporation from spending half a million dollars. They could have done ‘what if’ scenario’s in the model itself before actually investing in the hardware.

ROI for Creating a Performance Model

Building a performance model of an IT infrastructure or the application is a time consuming task, bottom line it costs money. So what is the return of investment of such an effort? Let’s do a simple calculation of a small organisation to start with. This is small e-commerce organisation , with a J2ee application running on 6 servers , 2 web server , 2 app server and 2 database server . The effort for performance modeling might be around 1 man months. With say 70 USD /hour billing rate (if outsourced to India it will cost half) it will cost the organisation US 6000. 6000 USD only to find out that they need to add few extra CPU in the application server or to increase the number of threads in the application server???? . Better could have bought two extra servers at this cost. Justification to the finance people for this budget will be tough. But it’s sure that blindly adding hardware will not improve performance, they definitely need consultancy what to add where to add, or it’s just a configuration issue. Are they ready for the Christmas rush how will they know? This performance modeling exercise is exactly this , with Performance model is a deliverable , with which “what if “ scenario’s can be simulated and analysedBut for large organisation ROI is quite high, as the cost of modelling will not increase linearly with the complexity of the infrastructure. And this cost can be reduced by outsourcing. Will you invest say 5000-20,000 US$ for this consultancy?? I think decision not whether you can effort it but can you effort not using this .