Sizing of hardware environment is usually done in cooperation with IBM or a business partner.
Work closely with the development team to hone the sizing estimation.
At this point in time, you should also consider which hardware platform you want to deploy and whether you prefer either scaling-up or scaling-out. The hardware platform decision is primarily dependent on your platform preference, which platforms have sizing information available, and which platforms are supported by WAS.
Scaling-up means to implement vertical scaling on a small number of machines with many processors. This can present fairly significant SPOF.
Scaling-out means using a larger number of smaller machines and is significantly more resilient, since it is unlikely that the failure of one small server is adequate to create a complete application outage. Scaling-out introduces a higher maintenance overhead.
An average response time is assumed for each transaction and calculations are performed to determine the estimated number of machines and processors the application will require. If your enterprise has a user experience team, they may have documented standards for typical response times that your new project will be required to meet.
For more accurate estimations of hardware requirements consider one of the benchmarking services offered by IBM or Business Partners.
Please notice that this sizing is for your production environment only. Based on this estimation, you not only have to update your production implementation design but you also have to update the designs for the integration and development environments accordingly. The key is to remember that changes to the production environment should be incorporated into the development and testing environments as well, if cost considerations do not merit otherwise.