Preparation work precedes the start of work on the first increment. [2] In Figure 17-2, the preparation work is similar to analyzing requirements and establishing the system architecture of the future system.
Lack of preparation work on the migration strategy and plan can increase the total cost of the system 32 percent. [3] The goal of the preparation phase is to reduce the uncertainty in key cost drivers: function and data migration, user interface, COTS software and hardware, and performance. The availability of tools for developing and verifying architectural specifications affects the efficiency of this phase, especially with a large legacy system.
Cost estimation for preparation work is performed using a cost model with the complete estimated modernized system size, estimated from the legacy code size and converted with the conversion ratio. With the system size and appropriate model inputs set to reflect the product, project, host platform, and people, an estimate can be produced for a big-bang ”single-increment development and deployment ”effort. The only part of this estimate we are interested in is the cost and schedule for the requirements and architecting phase, that is, only the portion of the estimate that applies to the preparation work. There are additional factors that influence the effort required in performing the preparation work. They are the structure of the legacy code, the availability of complete documentation that describes the internals of the legacy system, availability of accurate user's manuals, availability of internal/external interface documentation, and the availability of architects /programmers who are familiar with the legacy system. Assuming a complete estimated system size of 892 KSLOC ”estimated 713.6 KSLOC + 25 percent growth ”for RSS, relevant experience, availability of effective automated tools, and a compressed schedule, it could take from as many as 53 people for 17 months to as few as 27 people for 13 months to complete the preparation work. These numbers were calculated using the COCOMO II tool. Note that adjustments in the factors used by the formulas can result in widely varying estimates. |