Value-based Multiple Software Projects Scheduling with Genetic Algorithm Junchao Xiao, Qing Wang, Mingshu Li, Qiusong Yang, Lizi Xie, Dapeng Liu

Transcription

1 Value-based Multple Software Projects Schedulng wth Genetc Algorthm Junchao Xao, Qng Wang, Mngshu L, Qusong Yang, Lz Xe, Dapeng Lu Laboratory for Internet Software Technologes Insttute of Software, Chnese Academy of Scences

2 Agenda Introducton Motvatng Example Value Functon for Mult-project Schedulng Mult-project Schedulng wth Genetc Algorthm (GA) Case study Conclusons and Future Work 2

3 Agenda Introducton Motvatng Example Value Functon for Mult-project Schedulng Mult-project Schedulng wth Genetc Algorthm (GA) Case study Conclusons and Future Work 3

4 Background Mult-project envronment Contenton resource requrements among multple projects The projects may have dfferent stakeholders who bear dfferent requrements and preferences Each project holds dfferent constrants and dfferent value objectves One of the goals of an organzaton Acheve the maxmum value from the projects and response to the changng market tmely 4

5 Problems n mult-project schedulng Defne the value obtaned by schedulng accordng to constrants, value objectves and possble schedulng results n projects Provde a mult-project schedulng method whch can obtan the (near-) maxmum value for the organzaton Need decson support to managers 5

6 Related Work Mult-objectve release plannng Smulaton method Project portfolo management Resource schedulng n software projects 6

7 Our Method Value-based multple software projects schedulng method by usng a genetc algorthm Value functon n mult-project envronments s defned to gude the schedulng Genetc algorthm (GA) s adopted to tackle the problem of hgh complexty and can help the schedulng get nearly optmal solutons wth hgh effcency 7

8 Agenda Introducton Motvatng Example Value Functon for Mult-project Schedulng Mult-project Schedulng wth Genetc Algorthm (GA) Case study Conclusons and Future Work 8

9 RA1 P1 URA2 P2 RA3 AD1 AD3 IMP1a IMP1b WTC1 IMP2 DD3a DD3b DD3c COD3a COD3b COD3c TST1 TST2 TST3 Annotatons: RA Requrement Analyss; AD Archtecture Desgn; IMP Implementaton; WTC Wrte Test Case; TST Testng; URA Upgradng Requrement Analyss; DD Detaled Desgn; COD - Codng P3 WTC3 Beneft Increase n customer satsfacton More money earned by the organzaton Penalty Compensaton asked for by the customer Decrease n customer satsfacton 9

10 Schedule constrant P1 P2 P3 [ , ] [ , ] [ , ] Cost constrant 2*10 5 5* *10 5 Preference Cost preference Schedule preference Cost preference Schedule ahead beneft ($/day) Schedule postpone penalty ($/day) Cost saved beneft ($) Equal to saved Equal to saved Equal to saved Cost exceeded penalty ($) Equal to exceeded Equal to exceeded Equal to exceeded Project falure penalty($) Project mportance preference

11 Agenda Introducton Motvatng Example Value Functon for Mult-project Schedulng Mult-project Schedulng wth Genetc Algorthm (GA) Case study Conclusons and Future Work 11

12 Descrpton of Projects P = (ActSet, ConSet, PWSet ) Actvty Set: Each actvty s descrbed by the attrbutes ncludng dentfcaton (ID), type (TYPE), sze (SIZE), and requred sklls for human resources (SKLR) Constrant Set RA1 schedule constrant: [SD, DD] cost constrant Preference Weght Set preference weght of the project (PPW) preference weght of the schedule (SPW) and the cost (CPW) AD1 WTC1 IMP1a IMP1b TST1 12

13 Descrpton of Human Resources dentfcaton (ID) executable actvty type set (EATS) skll set (SKLS) experence data (EXPD) salary per man-hour (SALR) schedulable tme and workload (STMW) 13

14 Mult-project Value Functon Value of an organzaton Value of project P1 Value of project P2 Schedule value of P1 Cost value of P1 Schedule value of P2 Cost value of P2 14

15 Mult-project Value Functon Project Schedule Value (SValue) SBeneft CSB DD AFD 2 ( DD AFD ) SPenalty CSP AFD DD 2 ( AFD DD ) SValue SBeneft - SPenalty 15

16 Mult-project Value Functon Project Cost Value (CValue) CBeneft CCB CST APrjCST 2 ( CST APrjCST ) CPenalty CCP APrjCST CST 2 ( APrjCST CST ) CValue CBeneft - CPenalty 16

17 Mult-project Value Functon Project P Value CPW SPW CBeneft SBeneft CPenalty SPenalty Organzaton Value mult k PPW Value 1 17

18 Agenda Introducton Motvatng Example Value Functon for Mult-project Schedulng Mult-project Schedulng wth Genetc Algorthm (GA) Case study Conclusons and Future Work 18

19 Generate the ntal populaton accordng to the chromosome length and populaton scale Populaton evoluton Evaluate the ftness for each chromosome Select the chromosome for the next generaton on the bass of ftness cross and mutate thus generatng the next generaton Satsfy the expectng value or complete certan generatons Select the chromosome wth hghest ftness n the fnal generaton, and generate the fnal schedulng result 19

20 Structure of the Chromosome Encode A1, A2,, AN One Actvty Prorty One Capable Resource Code One Capable Resource Code Code Human resource genes Prorty genes A 1 A 2... A N Prorty for A 1... Prorty for A N 0/1 0/1... 0/1 0/1 0/1... 0/1... 0/1 0/1... 0/1 0/1... 0/1... 0/1... 0/1 HR 1,1 HR 1,2... HR 1,t1 HR 2,1 HR 2,2... HR 2,t2... HR N,1 HR N,2... HR N,tN Sze = g Sze = g 20

21 Structure of the Chromosome Decode (1) Select all the actvtes that do not have precedent actvtes or whose precedent actvtes have been assgned. If no such actvty exsts, then decodng s completed. (2) Sort all selected actvtes as a queue accordng to ther prorty from hgh to low. (3) For each actvty ACT n ths queue, do the followng steps: a) Set the capable human resources whose correspondng gene value s 1 as the scheduled human resources for ACT. b) Set the start date of ACT as the current date. c) Allocate all the schedulable workload of all the scheduled human resources n the current date to ACT and update the avalablty state of the resources. d) If the scheduled workload to ACT can complete ACT, then set current date be the due date of ACT and update the start date of the actvtes whose precedent actvty s ACT as the current date. Go to (3). e) Add one day to the current date, go to (c). (4) Go to (1). 21

22 Ftness Functon of the Chromosome Ftness Value mult 1 1 Value 2 mult 2 f Value mult 1 f Value mult [1,1] f Value mult 1 22

23 Runnng the Genetc Algorthm Set the parameters for runnng the GA Populaton scale (PS): the number of the chromosomes. Mutaton rate (MR): the possblty of mutaton to chromosome. Maxmum generaton number: the largest number of generatons. Termnaton condton: when the runnng of the GA should be termnated. After parameter settng, the schedulng wll be performed accordng to GA steps. 23

24 Agenda Introducton Motvatng Example Value Functon for Mult-project Schedulng Mult-project Schedulng wth Genetc Algorthm (GA) Case study Conclusons and Future Work 24

25 Projects and Human Resources RA1 P1 URA2 P2 RA3 AD1 AD3 IMP1a IMP1b WTC1 IMP2 DD3a DD3b DD3c COD3a COD3b COD3c TST1 TST2 TST3 Annotatons: RA Requrement Analyss; AD Archtecture Desgn; IMP Implementaton; WTC Wrte Test Case; TST Testng; URA Upgradng Requrement Analyss; DD Detaled Desgn; COD - Codng P3 WTC3 22 human resources Each human resource has stable productvty n each of the executable actvty types. 25

26 EATS EXPD(KLOC/Man-Hour) SALR HR1 RA P RA = HR2 RA P RA = HR3 RA P RA = HR4 RA P RA = HR5 RA P RA = HR6 AD, DD P AD = 0.06, P DD = HR7 AD, DD P AD = 0.055, P DD = HR8 AD, DD P AD = 0.05, P DD = HR9 AD, DD P AD = 0.04, P DD = HR10 IMP, DD, COD P IMP =0.025, P DD =0.05, P COD = HR11 IMP, DD, COD P IMP = 0.025, P DD = 0.05, P COD = HR12 IMP, DD, COD P IMP = 0.02, P DD = 0.05, P COD = HR13 IMP, DD, COD P IMP =0.02, P DD =0.03, P COD = HR14 IMP, DD, COD P IMP =0.02, P DD = 0.03, P COD = HR15 COD P COD = HR16 COD P COD = HR17 COD P COD = HR18 WTC, TST P WTC = 0.045, P TST = HR19 WTC, TST P WTC = 0.04, P TST = HR20 WTC, TST P WTC = 0.045, P TST = HR21 TST P TST = HR22 TST P TST =

27 TYPE SIZE PREA Capable Human Resource RA1 RA 25 No element exst HR1, HR2, HR3, HR4, HR5 AD1 AD 25 RA1 HR6, HR7, HR8, HR9 IMP1a IMP 10 AD1 HR10, HR11, HR12, HR13, HR14 IMP1b IMP 15 AD1 HR10, HR11, HR12, HR13, HR14 WTC1 WTC 25 RA1 HR18, HR19, HR20 TST1 TST 25 IMP1a,IMP1b,WTC1 HR18, HR19, HR20, HR21, HR22 URA2 RA 10 No element exst HR1, HR2, HR3, HR4, HR5 IMP2 IMP 10 URA2 HR10, HR11, HR12 TST2 TST 10 IMP2 HR18, HR19, HR20, HR21, HR22 RA3 RA 45 No element exst HR1, HR2, HR3, HR4, HR5 AD3 AD 45 RA3 HR6, HR7, HR8, HR9 DD3a DD 10 AD3 HR6,HR7,HR8,HR9,HR10,HR11,HR12,HR13,HR14 DD3b DD 20 AD3 HR6,HR7,HR8,HR9,HR10,HR11,HR12,HR13,HR14 DD3c DD 15 AD3 HR6,HR7,HR8,HR9,HR10,HR11,HR12,HR13,HR14 COD3a COD 10 DD3a HR10,HR11,HR12,HR13,HR14,HR15, HR16, HR17 COD3b COD 20 DD3b HR10,HR11,HR12,HR13,HR14,HR15, HR16, HR17 COD3c COD 15 DD3c HR10,HR11,HR12,HR13,HR14,HR15, HR16, HR17 WTC3 WTC 45 RA3 HR13, HR14, HR15 TST3 TST 45 COD3a, COD3b, COD3c, WTC3 HR13, HR14, HR15, HR16, HR17 The length of the capable human resource gene n the chromosome s

28 P1 P2 P3 [ , [ , [ , Schedule constrant ] ] ] Cost constrant 2*10 5 5* *10 5 Preference Cost preference Schedule preference Cost preference Schedule ahead beneft ($/day) Schedule postpone penalty ($/day) Cost saved beneft ($) Equal to saved Equal to saved Equal to saved Cost exceeded penalty ($) Equal to exceeded Equal to Equal to exceeded exceeded Project falure penalty($) Project mportance preference

29 Parameters of the Genetc Algorthm Populaton scale: 32. Prorty gene sze: 3, thus the length of the chromosome s: CL = *19 = 142. Mutaton rate: Maxmum generaton:

30 Organzaton Value Three smulaton runs of the algorthm Generaton 30

31 Project Value Value of 4*P1+P2 Project value affected by P3 s PPW P3 P3 P3 P3 P3 P3 P P1 P2 P3 4*P1+P Preference Weght of P3 31

32 Precedent Number of Date Saved Cost of P3 Precedent number of dates accordng to P3 s SPW P3 P3 P3 P3 P3 P3 P P P1 P2 P3 Saved Cost of P3 Schedule Weght of P

33 Beneft Dscussons A value functon s defned: t takes nto account the constrants and preferences of dfferent projects The schedulng results can reflect the value objectves of the organzaton: through the value functon, the schedulng results wll reflect the value objectves of the organzaton Provde the decson support for project managers: by settng dfferent coeffcents and preference weghts, project managers can compare the results of the resource schedulng easly 33

34 Agenda Introducton Motvatng Example Value Functon for Mult-project Schedulng Mult-project Schedulng wth Genetc Algorthm (GA) Case study Conclusons and Future Work 34

35 Conclusons The value functon takes full consderaton of the essental elements that affect the optmzng goal of schedulng such as schedule and cost. Based on ths value functon, the mult-project human resource schedulng method by usng a genetc algorthm s mplemented, whch allows the organzaton to obtan a near-maxmum value. Case study shows the method can take nto account the value objectves of the organzaton that uses ths method and effectvely reflect the organzaton value and provde decson support for managers. 35

36 Future Work Learnng curve of human resources Factors related to communcaton Overwork of human resources The comparson of GA wth other algorthms Analyss and justfcaton of GA parameters 36

37 Thank You! 37