U.S . Air Force S oftware Engineering Efficiency and Productivity for Information Operations Paul Braden U.S . Air Force CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 1
Agenda Introduction U.S . Air Force AFS C Way CMMI vs Agile Efficiency S tudy Definitions Tracking Productivity Definitions Tracking Conclusion CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 2
Introduction S oftware Engineer Problems Why are we here? 1/ 5/ 2002 https:/ / www.youtube.com/ watch? Chart Title 35 v=BKorP55Aqvg 30 6 25 20 5 15 4 1/ 9/ 2002 1/ 6/ 2002 10 5 3 0 2 1 0 Cat egory 1 Cat egory 2 Cat egory 3 Cat egory 4 1/ 8/ 2002 1/ 7/ 2002 S eries 1 S eries 2 S eries 3 S eries 1 S eries 2 CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 3
U.S Air Force AFS C Way “ S peed is GOOD” CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 4
U.S . Air Force AFS C Way The term “ S peed,” in the AFS C Way Lexicon, is meant to be synonymous with efficient processes that promote throughput paced to a Road to… goal. Efficient Productivity CMMI AGILE AGILE and CMMI CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 5
U.S . Air Force AFS C Way Important Definitions Road To… Networks Gates/ DBR Release Points Visual Displays S tandard Work (S cripting) Tools/ Tech Data Touch Time CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 6
U.S . Air Force AFS C Way Critical Path. A sequence of activities in a proj ect plan which must be completed by a specific time for the proj ect to be completed on its need date. CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 7
U.S . Air Force AFS C Way Little’s Law At st eady st at e, all product ion syst ems have an average t hroughput , work-in-progress (WIP), and flowt ime. The fundament al relat ionship bet ween all t hree is det ermined by Lit t le’s Law: WIP = t hroughput x flowt ime. CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 8
U.S . Air Force AFS C Way What led the Air Force to using the AFS C Way? CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 9
CMMI vs. Agile CMMI (Capability Maturity Agile (12 Principles for Model Integration) Proj ect Management) CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 10
CMMI vs. Agile Air Force uses CMMI as evaluation standard for software Agile is used as the process CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 11
CMMI vs. Agile Large scale S mall S cale Risk adverse S ingle team Mission critical Volatile requirements Management oversight S oftware only environment Reference: CMMI or Agile: Why Not Embrace Both! By Hillel Glazer, Jeff Dalton, David Anderson, Mike Konrad, S andy S hrum CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 12
Efficiency S tudy “ 100% relative efficiency is attained by any Decision Making Unit only when comparisons with other relevant DMUs do not provide evidence of inefficiency in the use of any input or output. As noted in the preface, this definition may be altered, expanded (or even replaced) when knowledge of 100% efficiency is available.” – 45S WI38-203 Most Efficienct Organization (MEO) CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 13
Efficiency S tudy Tinker AFB is CMMI level 3. Hill AFB is CMMI level 5. Warner Robins AFB is CMMI level 5. CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 14
Efficiency S tudy CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 15
Efficiency S tudy Tinker AFB looks at Return on Investment (ROI). Earned Value Management (EVM) looks at efficiency in terms of 1. cost, and 2. schedule. CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 16
Efficiency S tudy Tinker AFB 76 t h S MXG process flow was simulated to show inefficiencies and bottle necks S imple Tech Design Hardware Design Review Integration Implementation Manufacturing V and V 45 Days 34 Days 1 Day 120 Days 45 Days 95 Days 1 Day CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 17
Efficiency S tudy CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 18
Efficiency S tudy CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 19
Productivity S tudy IBM uses a computer software tool from CAS T to compare their code to industry best practices to ensure it is efficient and does not waste lines of code. Productivity rates are measured in function points per month or other time step. Quality is factored into this rate by checking for the defects and removal efficiency. Of the defects that are not removed, there is a further breakdown of high severity and low severity. Problem is finding ways to include time overruns and wasted time Biggest downfall to productivity is paperwork and admin Complexity is a factor that can normalize the productivity charts while removing some of the Lines of Code issues Introducing an idea of burden percentage to normalize cost for productivity CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 20
Productivity S tudy Air Force uses lines of code for productivity studies Technical Packages of S oftware (TPS ) are divided into S imple, Average, Complex, and Very Complex. Number of Flowdays are: S imple -> 211 Average -> 296 Complex -> 383 Very Complex -> 443 CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 21
Productivity One Way Analysis of Variance (ANOVA) tests showed that complexity factors were not good ways to differentiate the different TPS ’s. Best workers were put on Very Complex proj ects and were able to finish faster than S imple proj ects Only Average and Complex TPS had any differentiation CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 22
Conclusion Efficiency is dependent on Most Efficient Organization Productivity depends more on quality of programmer than on complexity Air Force uses efficiency and productivity in the Agile and CMMI constructs as measurements of success CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 23
Questions CLAS S IFICATION: UNCLAS S IFIED 13 May 2016 24
Recommend
More recommend
