Practical Implementation of COSYSMO Reuse Extension Gan Wang, Aaron Ankrum, Cort Millar, Alex Shernoff, Ricardo Valerdi ©2006 BAE Systems. Agenda • • • • Background Industry Definition BAE Systems Implementation Lessons Learned and Recommendations What Is the Problem? • Everything has a legacy • In particular, most of the systems we build and programs we execute are based on legacy systems, architecture, and experience – Rarely are we building something completely new from the “clean slate”. • Academic COSYSMO assumes all size drivers (Req, Int, Alg, Scn) are new, which is inadequate in estimating the systems we build today • Therefore, we must consider the impact of reuse on systems engineering size • We anticipate this will improve the model’s accuracy since reuse plays a significant role in today’s systems COSYSMO at BAE Systems • BAE Systems E&IS OG has developed an internal systems estimating tool based on COSYSMO – Collected 55 past program data points across lines of business and major sites – Developed platform-specific calibrations • Problem encountered: Academic COSYSMO could not provide the correlation and convergence required to create meaningful calibrations – One of the big issues was driver reuse • We worked with key stakeholders and developed the industry definition of COSYSMO Reuse • We developed an internal implementation/instantiation of the reuse extension Industry Definition Reuse Categories 1. New: Items that are completely new. 2. Adopted: Items that are incorporated unmodified. 3. Modified: Items that are reused but are tailored. 4. Deleted: Items that are removed from a system. Driver Counting Framework Difficult # System Requirements # System Interfaces Nominal # System Algorithms # Operational Scenarios Easy New Adopted Modified Deleted Application of Reuse at BAE Systems Implementation Strategy • Be consistent with industry definition – By further instantiating and refining the verbiage so as to avoid any potential conflicts and inconsistencies • Provide a clear and consistent operational guideline for driver counting and classification – By defining unambiguous verbiage of the reuse definition • Establish clear boundaries between categories to ensure easy separation and consistency – Steps rather than ramps – Delineated by common concepts Approach • Define two (orthogonal) dimensions of classification framework to enable finer grain estimation of the size drivers: – Reuse by Process: SE activities – Level of Difficulty by Relative Effort: easy, nominal, difficult Reuse Categories at BAE Systems 1. New: Items that are completely new 2. Managed: Items that are incorporated and require no added SE effort other than technical management 3. Adopted: Items that are incorporated unmodified but require verification and validation 4. Modified: Items that are incorporated but require tailoring or interface changes, and verification and validation 5. Deleted: Items that are removed from a legacy system, which require design analysis, tailoring or interface changes, and verification and validation Notes: • New items are generally unprecedented • Those items that are inherited but require architecture or implementation changes should be counted as New Added Categories “Managed” • We found, in particular, significant number of requirements require “technical management” level of effort only – They are still part of the system deliverables – They are still to be “sold off” to customers – They are not free • Applied for – Proven elements from previous iterations/spirals – Subcontract/teammate provided turn-key components Extended COSYSMO Relationship E PM NS A wr ( we,k e ,k wn ,k n ,k wd ,k d ,k ) CEM k r Where: PMNS = effort in Person Hours/Months (Nominal Schedule) A = calibration constant derived from historical project data k = {REQ, IF, ALG, SCN} r = {New, Managed, Adopted, Modified, Deleted} wr = weight for defined levels of size driver reuse wx = weight for “easy”, “nominal”, or “difficult” size driver Фx = quantity of “k” size driver E = represents diseconomy of scale CEM = composite effort multiplier Driver Classification Guideline • Separate the two dimensions: Apply reuse in terms of Process (activity) Apply levels of difficulty in term of Relative Effort Order of Classification New Easy Nominal Difficult Managed Adopted Modified First Pass Second Pass Deleted Define Reuse Weights Based on Activities Data Before Reuse Application Same Data After Reuse Application Excellent Correlation Achieved for Similar Programs PRED(30) PRED(25) PRED(20) Application Lessons Learned • Modified category, in particular, covers a wide range • The intent of the modified category is to capture those elements that involve tailoring changes only, with no changes to the internal architecture • If you modify everything, you should classify it as New • Consistency is the key – Between data points – Between calibration data and new estimate • What do we do for consistency? – Asked for a sample nominal requirement – Compare the productivity metric with peer programs – Graphical aids are helpful, e.g., Recommendations • Reuse weights are meant to be defined ones only. Once they are defined, you should stick with it and apply to all programs – If you have to change them, apply to all program data • Recommendation: industry should have a recommended value range for all reuse categories – To better communicate BOEs Contact Gan Wang gan.wang@baesystems.com 703-668-4259