assignment for kim woods

2 The M yth ic a l M an-M onth 2 The M yth ic a l M an-M onth Good co okin g fa kes tim e. If y o u are m ade to w ait, it is to s e rv e yo u bette r, and to p le a se yo u. M EN U OF R EST A U RAN T A N TO IN E. NEW O RLE AN S 13 14 The M yth ic a l M an -M on th M ore s o ftw are p ro je cts h av e g one a w ry f o r l a ck o f c ale n dar t im e t h an f o r a ll o th er c au se s c o m bin ed . W hy is th is c au se o f d is a ste r s o c o m mon?

F ir s t, o ur t e ch niq ues o f e stim atin g a re p oorly d ev elo ped . M ore s e rio usly , th ey r e fle ct a n u nvoic ed a ssu m ptio n w hic h i s q uite u n - t r u e, i.e ., th at a ll w ill g o w ell.

S eco nd, o ur e stim atin g t e ch niq ues f a lla cio usly c o nfu se e ffo rt w it h p ro gre ss, h id in g th e a ssu m ptio n th at m en a n d m onth s a re i n te rc h an geab le .

T hir d , b ecau se w e a re u ncerta in o f o ur e stim ate s, so ftw are m an ag ers o fte n l a ck t h e c o urte o us s tu bborn ness o f A nto in e's c h ef.

F ourth , sc h ed ule p ro gre ss is p oorly m onito re d . T ech niq ues p ro ven a n d r o utin e in o th er e n gin eerin g d is c ip lin es a re c o nsid ere d r a d ic al in nov atio ns in s o ftw are e n gin eerin g.

F if th , w hen s c h ed ule s lip pag e is r e co gniz ed , th e n atu ra l ( a n d t r a d itio nal) re sp onse i s t o a d d m an pow er. L ik e d ousin g a f ir e w ith g aso lin e, th is m ak es m atte rs w ors e , m uch w ors e . M ore fir e re - q uir e s m ore g aso lin e, a n d th us beg in s a r e g en era tiv e c y cle w hic h e n ds in d is a ste r.

S ch ed ule m onito rin g w ill b e th e s u bje ct o f a se p ara te e ssa y .

L et u s c o nsid er o th er a sp ects o f th e p ro ble m in m ore d eta il.

O ptim is m A ll p ro gra m mers a re o ptim is ts . P erh ap s t h is m odern s o rc ery e sp e- c ia lly a ttr a cts th ose w ho b elie v e in h ap py e n din gs a n d f a ir y g od- m oth ers . P erh ap s t h e h undre d s o f n itty f ru str a tio ns d riv e a w ay a ll b u t th ose w ho h ab itu ally fo cu s o n th e en d g oal. P erh ap s it is m ere ly t h at co m pute rs a re y oung, p ro gra m mers a re y ounger, a n d t h e y oung a re a lw ay s o ptim is ts . B ut h ow ev er t h e s e le ctio n p ro cess w ork s, th e r e su lt is in dis p uta b le : " T his tim e it w ill s u re ly r u n," o r " I ju st f o und th e la st b ug." S o t h e f ir s t f a ls e assu m ptio n th at u nderlie s th e s c h ed ulin g o f sy ste m spro gra m min g i s th at a ll w ill g o w ell, i.e ., th at e a ch ta sk w ill h ik e o nly a s lo ng a s it "o ught" to ta ke. Optim is m 15 T he p erv asiv en ess o f o ptim is m a m ong p ro gra m mers d ese rv es m ore th an a f lip a n aly sis . D oro th y S ay ers , in h er e x celle n t b ook, T he M in d o f t h e M aker, d iv id es c re ativ e a ctiv ity i n to th re e s ta g es:

t h e id ea, th e im ple m en ta tio n, a n d th e in te ra ctio n. A b ook, th en , o r a c o m pute r, o r a p ro gra m c o m es in to ex is te n ce f ir s t a s a n id ea l c o nstr u ct, b uilt o uts id e tim e a n d s p ace, b ut c o m ple te in th e m in d o f th e a u th or. It is re aliz ed in tim e a n d sp ace, b y p en , in k, a n d p ap er, o r b y w ir e , silic o n , a n d fe rrite . T he c re atio n is c o m ple te w hen so m eo ne re ad s th e b ook, u se s th e c o m pute r, o r ru ns th e p ro gra m , th ere b y in te ra ctin g w ith th e m in d o f th e m ak er.

T his d esc rip tio n, w hic h M is s S ay ers u se s to illu m in ate n ot o nly h um an c re ativ e a ctiv ity b ut a ls o t h e C hris tia n d octr in e o f t h e T rin ity , w ill h elp u s i n o ur p re se n t t a sk . F or t h e h um an m ak ers o f t h in gs, th e in co m ple te n esse s an d in co nsis te n cie s of our id eas b eco m e c le ar o nly d urin g i m ple m en ta tio n. T hus i t i s th at w ritin g, ex perim en ta tio n, " w ork in g o ut" a re e sse n tia l d is c ip lin es f o r th e th eo re tic ia n .

I n m an y c re ativ e a ctiv itie s t h e m ed iu m o f e x ecu tio n i s i n tr a ct- a b le . L um ber sp lits ; p ain ts sm ear; e le ctr ic al c ir c u its rin g. T hese p hysic al lim ita tio ns o f th e m ed iu m c o nstr a in th e id eas th at m ay b e e x pre sse d , a n d th ey als o c re ate u nex pecte d d if fic u ltie s in th e i m ple m en ta tio n.

I m ple m en ta tio n, th en , ta k es tim e a n d s w eat b oth becau se o f t h e p hysic al m ed ia a n d b ec au se o f th e in ad eq uacie s o f th e u nder- l y in g i d eas. W e te n d to b la m e th e p hysic al m ed ia f o r m ost o f o ur i m ple m en ta tio n d if fic u ltie s; fo r th e m ed ia a re n ot "o urs " in th e w ay th e id eas a re , a n d o ur p rid e c o lo rs o ur ju dgm en t.

C om pute r pro gra m min g, h ow ev er, cre ate s w ith an ex ceed - i n gly tr a cta b le m ed iu m . The pro gra m mer build s fro m pu re t h ought- s tu ff: c o ncep ts a n d v ery f le x ib le re p re se n ta tio ns th ere o f.

B ecau se th e m ed iu m is tr a cta b le , w e e x pect fe w d if fic u ltie s in i m ple m en ta tio n; h en ce o ur p erv asiv e o ptim is m . B ecau se o ur i d eas a re f a u lty , w e h av e b ugs; h en ce o ur o ptim is m is u nju stif ie d .

I n a sin gle ta sk , th e a ssu m ptio n th at a ll w ill g o w ell h as a p ro bab ilis tic e ffe ct o n t h e s c h ed ule . I t m ig htin deed g o a s 16 The M yth ic a l M an -M on th f o r th ere is a p ro bab ility d is tr ib utio n fo r th e d ela y th at w ill b e e n co unte re d , a n d " n o d ela y " h as a f in ite p ro bab ility . A la rg e p ro - g ra m min g e ffo rt, how ev er, c o nsis ts o f m an y ta sk s, s o m e c h ain ed e n d-to -e n d. T he p ro bab ility th at e ach w ill g o w ell b eco m es v an - i s h in gly sm all.

T he'M an -M on th T he s e co nd f a lla cio us t h ought m ode i s e x pre sse d i n t h e v ery unit o f e ffo rt u se d i n e stim atin g a n d s c h ed ulin g: t h e m an -m onth . C ost d oes in deed v ary a s th e p ro duct o f th e n um ber o f m en a n d th e n um ber o f m onth s. P ro gre ss d oes n ot. H en ce t h e m an-m onth a s a u nit fo r m ea su rin g th e siz e o f a jo b is a d angero us a nd d ecep tiv e m yth . It im plie s th at m en a n d m onth s a re in te rc h an geab le .

M en a n d m onth s a re i n te rc h an geab le c o m moditie s o nly w hen a t a sk c an b e p artitio ned a m ong m an y w ork ers w ith n o c o m munic a - t io n a m ong t h em ( F ig . 2 .1 ). T his i s t r u e o f r e ap in g w heat o r p ic k in g c o tto n; i t i s n ot e v en a p pro xim ate ly t r u e o f s y ste m s p ro gra m min g. M en F ig . 2 .1 Tim e v ers u s n um ber o f w ork ers — perfe ctly p artitio nab le ta sk The M an -M onth 17 W hen a ta sk c an not b e p artitio ned b ecau se o f s e q uen tia l c o n- s tr a in ts , th e a p plic atio n o f m ore e ffo rt h as n o e ffe ct o n t h e s c h ed - u le ( F ig . 2 .2 ). T he b earin g o f a c h ild ta k es n in e m onth s, n o m atte r h ow m an y w om en a re a ssig ned . M an y s o ftw are ta sk s h av e th is c h ara cte ris tic b ecau se o f th e s e q uen tia l n atu re o f d eb uggin g. Fig . 2 .2 Tim e v ers u s n um ber o f w ork ers — unpartitio nab le ta sk In t a sk s t h at c an b e p artitio ned b ut w hic h r e q uir e c o m munic a- t io n a m ong th e su bta sk s, th e e ffo rt o f c o m munic atio n m ust b e a d ded to th e a m ount o f w ork to b e d one. T here fo re t h e b est th at c an b e d one is s o m ew hat p oore r th an a n e v en tr a d e o f m en f o r m onth s ( F ig . 2 .3 ). 18 The M yth ic a l M an -M on th M en F ig . 2.3 Tim e v ers u s n um ber o f w ork ers — partitio nab le ta sk re q uir in g c o m munic atio n T he a d ded b urd en o f c o m munic atio n is m ad e u p o f t w o p arts , t r a in in g a n d in te rc o m munic atio n. E ach w ork er m ust b e tr a in ed in t h e t e ch nolo gy, t h e g oals o f t h e e ffo rt, t h e o vera ll s tr a te g y, a n d t h e p la n o f w ork . T his tr a in in g c an not b e p artitio ned , s o th is p art o f t h e a d ded e ffo rt v arie s lin early w ith th e n um ber o f w ork ers . 1 In te rc o m munic atio n is w ors e . If e ach p art o f t h e ta sk m ust b e s e p ara te ly c o ord in ate d w ith e ach o th er p art/ th e e ffo rt in cre ase s a s n (n -I)/2 . T hre e w ork ers re q uir e th re e tim es as m uch p air w is e i n te rc o m munic atio n a s tw o; f o ur re q uir e s ix tim es a s m uch a s tw o.

I f, m ore o ver, t h ere n eed t o b e c o nfe re n ces a m ong t h re e, f o ur, e tc ., w ork ers t o r e so lv e t h in gs j o in tly , m atte rs g et w ors e y et. T he a d ded e ffo rt o f c o m mun ic atin g m ay fu lly c o unte ra ct th e d iv is io n o f th e orig in al ta sk a n d b rin g u s to th e s itu atio n o f F ig . 2 .4 . Syste m s T est 19 M en F ig . 2 .4 Tim e v ers u s n um ber o f w ork ers — ta sk w ith c o m ple x i n te rre la - tio nsh ip s S in ce s o ftw are c o nstr u ctio n i s i n here n tly a s y ste m s e ffo rt— an e x erc is e in c o m ple x in te rr e la tio nsh ip s— co m munic atio n e ffo rt is g re at, a n d i t q uic k ly d om in ate s t h e d ecre ase i n i n div id ual t a sk tim e b ro ught a b out b y p artitio nin g. A ddin g m ore m en t h en l e n gth en s, n ot s h orte n s, th e s c h ed ule .

S yste m s T est N o p arts o f t h e s c h ed ule a re s o t h oro ughly a ffe cte d b y s e q uen tia l c o nstr a in ts as co m ponen t d eb uggin g an d sy ste m te st. F urth er- m ore , th e tim e r e q uir e d d ep en ds o n th e n um ber a n d s u btle ty o f t h e e rro rs e n co unte re d . T heo re tic ally th is n um ber s h ould b e z ero .

B ecau se o f o ptim is m , w e u su ally e x pect t h e n um ber o f b ugs t o b e 20 The M yth ic a l M an -M onth s m alle r th an it tu rn s o ut to b e. T here fo re te stin g is u su ally th e m ost m is -s c h ed ule d p art o f p ro gra m min g.

F or s o m e y ears I h av e b een s u ccessfu lly u sin g t h e f o llo w in g r u le o f th um b fo r sc h ed ulin g a so ftw are ta sk : l/3 p la n nin gl/6 c o din gl/4 c o m pon en t te st a n d e arly s y ste m te stl/4 s y ste m t e st, a ll c o m ponen ts i n h an d.

T his d if fe rs f ro m c o nven tio nal s c h ed ulin g i n s e v era l i m porta n t w ay s:

1 . T he f ra ctio n d ev ote d t o p la n nin g i s l a rg er t h an n orm al. E ven s o , it is b are ly e n ough to p ro duce a d eta ile d a n d s o lid s p ecif i- c atio n, a n d n ot e n ough to in clu de r e se arc h o r e x plo ra tio n o f t o ta lly n ew te ch niq ues.

2 . T he h alf o f t h e s c h ed ule d ev ote d to d eb uggin g o f c o m ple te d c o de is m uch la rg er th an n orm al.

3 . T he p art th at is e asy to e stim ate , i.e ., c o din g, is g iv en o nly o ne-s ix th o f th e sc h ed ule .

I n e x am in in g c o nven tio nally s c h ed ule d p ro je cts , I h av e f o u nd t h at fe w a llo w ed o ne-h alf o f th e p ro je cte d sc h ed ule fo r te stin g, b ut t h at m ost d id in deed s p en d h alf o f t h e a ctu al s c h ed ule f o r t h at p urp ose . M an y o f th ese w ere o n sc h ed ule u ntil an d ex cep t in s y ste m te stin g. 2 F ailu re to a llo w e n ough tim e f o r s y ste m te st, in p artic u la r, is p ecu lia rly d is a str o us. S in ce th e d ela y c o m es at th e e n d o f th e s c h ed ule , n o o ne is a w are o f sc h ed ule tr o uble u ntil a lm ost th e d eliv ery d ate . B ad n ew s, la te a n d w ith out w arn in g, is u nse ttlin g t o c u sto m ers a n d to m an ag ers .

F urth erm ore , dela y a t th is p oin t h as u nusu ally se v ere fin an - c ia l, a s w ell a s p sy ch olo gic al, re p erc u ssio ns. T he p ro je ct is f u lly s ta ffe d , a n d c o st- p er-d ay i s m ax im um . M ore s e rio usly , t h e s o ft- w are is to s u pp ort o th er b usin ess e ffo rt ( s h ip pin g o f c o m pute rs , o pera tio n o f n ew f a c ilitie s, e tc .) a n d t h e s e co ndary c o sts o f d ela y - i n g t h ese a re v ery h ig h, f o r i t i s a lm ost t im e f o r s o ftw are s h ip m en t. Reg en era tiv e S ch ed ule D is a ste r 21 In deed , th ese s e co ndary c o sts m ay f a r o utw eig h a ll o th ers . It is t h ere fo re v ery im porta n t to a llo w e n ough s y ste m te st t im e in th e o rig in al sc h ed ule .

G utle ss E stim atin g O bse rv e t h at f o r t h e p ro gra m mer, a s f o r t h e c h ef, t h e u rg en cy o f t h e p atr o n m ay g overn th e s c h ed ule d c o m ple tio n o f th e ta sk , b ut i t c an not g overn th e a ctu al c o m ple tio n. A n o m ele tte , p ro m is e d i n t w o m in ute s, m ay a p pear t o b e p ro gre ssin g n ic ely . B ut w hen i t h as n ot s e t i n t w o m in ute s, t h e c u sto m er h as t w o c h oic esÐ wait o r e at i t r a w . S oftw are c u sto m ers h av e h ad th e s a m e c h oic es.

T he c o ok h as a n oth er c h oic e; h e c an tu rn u p th e h eat. T he r e su lt is o fte n a n o m ele tte n oth in g c an s a v eÐ burn ed in o ne p art, r a w in a n oth er.

N ow I d o n ot th in k so ftw are m an ag ers h av e le ss in here n t c o ura g e a n d f ir m ness t h an c h efs , n or t h an oth er e n gin eerin g m an - a g ers . B ut fa ls e s c h ed uli n g to m atc h th e p atr o n's d esir e d d ate is m uch m ore c o m mon i n o ur d is c ip lin e th an els e w here in e n gin eer- i n g. It is v ery d if fic u lt to m ak e a v ig oro us, p la u sib le , a n d jo b- r is k in g d efe n se o f a n e stim ate th at is d eriv ed b y n o q uan tita tiv e m eth od, su pporte d by li ttle data , an d certif ie d ch ie fly by th e h unch es o f th e m an ag ers .

C le arly tw o so lu tio ns a re n eed ed . W e n eed to d ev elo p a n d p ublic iz e p ro ductiv ity fig ure s, b ug-in cid en ce fig ure s, estim atin g r u le s, a n d s o o n. T he w hole p ro f e ssio n c an o nly p ro fit f ro m s h arin g s u ch d ata .

U ntil estim atin g is o n a so under b asis , in div id ual m an ag ers w ill n eed to stif fe n th eir b ack bones a n d d efe n d th eir e stim ate s w ith th e a ssu ra n ce th at th eir p oor h unch es a re b ette r th an w is h - d eriv ed estim ate s.

R eg en era tiv e S ch ed ule D is a ste r W hat does o ne d o w hen a n e sse n tia l s o ftw are p ro je ct is b eh in d s c h ed ule ? A dd m an pow er, n atu ra lly . A s F ig s. 2 .1 t h ro ugh 2 .4 s u g- g est, th is m ay o r m ay n o t h elp . 22 The M yth ic al M an -M onth L et u s c o nsid er a n e x am ple . 3 S uppose a t a sk i s e stim ate d a t 1 2 m an -m onth s a n d a ssig ned t o t h re e m en f o r f o u r m onth s, a n d t h at t h ere a re m easu ra b le m ile p osts A , B , C , D , w hic h a re s c h ed ule d to f a ll a t th e e n d o f e ac h m onth (F ig . 2 .5 ). N ow su ppose th e fir s t m ile p ost is not re ach ed u ntil tw o m onth s h av e e la p se d (F ig . 2 .6 ). W hat a re th e a lte rn ativ es f a cin g t h e m an ag er?

1 . A ssu m e t h at t h e t a sk m ust b e d one o n t im e. A ssu m e t h at only t h e f ir s t p art o f t h e t a sk w as m is e stim ate d , s o F ig . 2 .6 t e lls t h e s to ry a ccu ra te ly . T hen 9 m an -m onth s o f e ffo rt r e m ain , a n d tw o m onth s, s o 4 V £ m en w ill b e n eed ed . A dd 2 m en t o t h e 3 a ssig ned .

2 . A ssu m e t h at t h e t a sk m ust b e d one o n t im e. A ssu m e t h at t h e w hole e stim ate w as u nif o rm ly lo w , so th at F ig . 2 .7 re ally d esc rib es t h e s itu atio n. T hen 1 8 m an -m onth s o f e ffo rt r e m ain , a n d t w o m onth s, s o 9 m en w ill b e n eed ed . A dd 6 m en to t h e 3 a ssig ned . F ig ure 2 .5 Reg en era tiv e S ch ed ule D is a ste r 23 F ig ure 2 ,6 F ig ure 2 .7 24 The M yth ic a l M an -M on th 3 . R esc h ed ule . I l ik e t h e a d vic e g iv en b y P . F ag g, a n e x perie n ced h ard w are en gin eer, "T ak e no sm all slip s." T hat is , allo w e n ough tim e in th e n ew s c h ed ule to e n su re th at th e w ork c an b e c are fu lly a n d th oro ughly d one, a n d th at r e sc h ed ulin g w ill n ot h av e to b e d one a g ain .

4 . T rim th e ta sk . I n p ra ctic e th is te n ds to h ap pen a n yw ay , o nce t h e te am o bse rv es sc h ed ule slip pag e. W here th e se co ndary c o sts o f d ela y a re v ery h ig h, th is is th e o nly fe asib le a ctio n.

T he m an ag er's o nly a lte rn ativ es a re to tr im it fo rm ally a n d c are fu lly , to re sc h ed ule , or to w atc h th e ta sk g et sile n tly t r im med b y h asty d esig n a n d in co m ple te te stin g.

I n th e fir s t tw o case s, in sis tin g th at th e u nalte re d ta sk b e c o m ple te d i n f o u r m onth s is d is a str o us. C onsid er t h e r e g en era tiv e e ffe cts , f o r e x am ple , f o r t h e f ir s t a lte rn ativ e ( F ig . 2 .8 ). T he t w o n ew m en , h ow ev er c o m pete n t a n d h ow ev er q uic k ly r e cru ite d , w ill r e - q uir e tr a in in g in th e ta sk b y o ne o f th e e x perie n ced m en . If th is t a k es a m onth , 3 m an-m onth s w ill h ave b een d evo te d t o w ork n ot i n th e orig in al e stim ate . F urth erm ore , t h e t a sk , o rig in ally p artitio ned t h re e w ay s, m ust b e re p artitio ned in to fiv e p arts ; h en ce so m e w ork a lr e ad y d one w il l b e lo st, a n d s y ste m te stin g m ust b e le n gth en ed .

S o a t th e e n d o f th e th ir d m onth , s u bsta n tia lly m ore th an 7 m an - m onth s o f e ffo rt r e m ain , a n d 5 tr a in ed p eo ple a n d o ne m onth a re a v aila b le . A s F ig . 2 .8 s u ggests , th e p ro duct is ju st a s la te a s if n o o ne h ad b een a d ded (F ig . 2 .6 ).

T o h op e t o g et d on e i n f o ur m onth s, c o nsid erin g o nly t r a in in g t im e a n d n ot r e p artitio nin g a n d e x tr a s y ste m s te st, w ould re q uir e a d din g 4 m en , n ot 2 , a t th e e n d o f th e se co nd m onth . T o c o ver r e p artitio nin g a n d s y ste m te st e ffe cts , o ne w ould h av e to a d d s till o th er m en . N ow , h ow ev er, o ne h as a t le ast a 7 -m an te am , n ot a 3 -m an o ne; th us s u ch a sp ects a s te am o rg an iz atio n a n d ta sk d iv i- s io n a re d if fe re n t in k in d, n ot m ere ly in d eg re e.

N otic e th at b y th e e n d o f th e th ir d month th in gs lo ok v ery b la ck . T he M arc h 1 m ile sto ne h as n ot b een r e ach ed i n s p ite o f a ll Reg en era tiv e S ch ed ule D is a ste r 25 th e m an ag eria l e ffo rt. T he t e m pta tio n i s v ery s tr o ng t o r e p eat t h e c y cle , a d din g y et m ore m an pow er. T here in lie s m ad ness.

T he fo re g oin g assu m ed th at only th e fir s t m ile sto ne w as m is e stim ate d . If o n M arc h I o ne m ak es th e c o nse rv ativ e a ssu m p- t io n t h at t h e w hole s c h ed ule w as o ptim is tic , a s F ig . 2 .7 d ep ic ts , o ne w an ts to a d d 6 m en ju st to th e o rig in al ta sk . C alc u la tio n o f th e t r a in in g, r e p artitio nin g, s y ste m t e stin g e ffe cts i s l e ft a s a n e x erc is e f o r th e re ad er. W ith out a d oubt, th e re g en era tiv e d is a ste r w ill y ie ld a p oore r p ro duct, la te r, th an w ould re sc h ed ulin g w ith th e o rig in al th re e m en , u nau g m en te d .

O vers im plif y in g o utr a g eo usly , w e s ta te B ro oks's L aw : Addin g m anpow er to a la te s o ftw are p ro je ct m akes it la te r.

T his th en is th e d em yth olo giz in g o f th e m an -m onth . T he n um ber o f m onth s o f a p ro je ct d ep en ds u pon its s e q uen tia l c o n- F ig u re 2 .8 26 The M yth ic al M an -M onth s tr a in ts . T he m ax im um n um ber o f m en d ep en ds u pon t h e n um ber o f i n dep en den t s u bta sk s. F ro m t h ese t w o q uan titie s o ne c an d eriv e s c h ed ule s u sin g fe w er m en a n d m ore m onth s. (T he o nly ris k is p ro duct o bso le sc en ce.) O ne c an not, h ow ev er, g et w ork ab le s c h ed - u le s u sin g m ore m en a n d f e w er m onth s. M ore s o ftw are p ro je cts h av e g one a w ry f o r l a ck o f c ale n dar t im e t h an f o r a ll o th er c au se s co m bin ed .