Software polabiliy. Reliability = Probabilihy that a Combonent coil ber form it's intended function satisfactorily for a given reriog off time (l) vith given cyeumstances « Software goes evough same phases of development- ue Raq uivement Analysis » Design 7 Manufactre/ coding > Jest > Packaging v Ais pateh Telineky ~ Movionda ( | Aesumphions |- . b Each of N faults in software will cause a failure after a kime which is distributed exponentially. ! Tw oT a fe 1-7 - 8 byte, £00 exponential ia ? Saftware faults ave independent of each other and eaprally likely fo cause failure Bp Failure, vale at any Hore i proportional to cuvvent faut content of soffuare systemlum oe, ee Pa. A defected fault is coppected with senky in 4 } negligible amount of time cud no new faults ave intyoduced 7 Nh -— fire 38 == vale Se 5 Tie en e+ [N= eal ge + WW -trayh $lti) 2 HN se W-tiyd 2. ae n = f(y) ha ~ x ae ote eee 7 2 Pt —@ alin > oe. 7 = z (N-tad) te —@ now, Chal) _ oy $= — -@ 2 (tad) te cyfs ov, >= Movawda Modet [4945]. — [Heston | 41> Fach of rail cause the tn finite oo of faults in a sof twave an Corbuve after atime, Ahidh is faut = 4] system aigh ibuted exponentially oD Here # Coftware foulls are independent 49 each other Failuve vale deceeaces geomeltteally aften gach facil & detected and Cory ected 4 A detected foult ts corvecte 7 negigirl’ amount of tiene x a oa ¥ S 4 with eer tainity, in 6 aud no new faults are (Oko= —ducef aGked 1a “ |b unknown ot CStart facture al jp estimated -filuve pe vale as , _—- ohh a Le f 1 f(y , ter ua 7 —bK dy fs DW 8 i n el pe TS re ee ie i 0 a “ ft bet = dab 2 Ped + 7 UHty Ak — DDK treo | ta te uy a » Ake L) Wey Pray Zw 72 bey Ha. I m6 ig from (9 pra Nn j- Ke ti @ isd bom Gg, 4 ioe n to k 2 Chay Fon Lek seeDiscrete Frarne voy «| = We have said, M ng of Fest cases, ay Take a test case Crandornly » Run the sof-tuave test case. ? Nole down the ouk come, 7 Success 3 Not geting the evpecte Failure outcen. N [orath Sat is bug) expel cuane ? FepCat The proces abe Sutiome . Tpke ‘Gof taave Releace fimd ~> — When should we stop We (My, Wy Te, ibe eh Probability aistribution of He buknown no of UMdiscovered bug. Cremainingy My = probability ‘Thot the software has w remnaining bugs Ti( hb) = Priav distr buhon of remaiving 00. of bugs after having a histovy h, and F test casecave lected pee t T(o,0) > history | © fest case ave testedflues & Ristovy SDR Sd sequance of cuccess awd failure. failure Sees Ost, 2). or ee 79 cue Ti(o, 0) 3 wCh4) S > Ths, bth) _ Sa Ths bay) > T(hef, ty £ CT (hh bat) ete) Pe TAR, Ee), " Wht) Assumph ons. 1> Zn this disevele frame cork, one can g2k amos One bug by testing one fect case- 4 2 re ar 3 ss softusave faile fault is Fmmediately detected and fixed without incuvvivg any Ul a g ass, ? Tf weg follow ckep @ ind oo Fixed the deheeted eg twmunediately we gel 6 ren of one unit > fag eel for esting each case is assumed bo be same aud equals 1 ©. This Sa The one step posterior Probability of im no of Undis dace, bugs can bu written ay Th, (hs b44) = Tos Cyt) Pay |. =]. z 1; (ht) By Ps canditional robabil A) obability ¢ : we} tn&ecna) tad f See ese OMe" that theve VLA tp] Th Che ter) © Me Wut) an 2 Wega 0 Peo) S uy OL, 01 U4 0.05 O05 ods ods 09 oa Gs th oo Bs 2 a Ths t oo) An. Mra, he) P Ty Cs, be 4) ee! 2 # Me Cyt): Py vr0 es (Lo.5)x 0.02 ea +4 %0-02 Ty Chy-tasy > GLAM OS X0-02, OE 7 7 OS Koo, =~ O2 Ty (hs, tery OOK OM Lo exper PO = Tyla, = Ms Che tay, =U] eho hoFT (y,trt) = (as, 3,04, 04, 04 98.94, 99, n ow, A 0 TChes tay 2 22% a0 = 02S BahK | Ceo.2y % 970) 0280: 02 TM Uns, bar) 2 20. Ons, Ean) 0.8 x0. 0, Ty Uns, Eta) 2 O48 0-0), 20-1952 M3 Ch OR 0.04 age) “= My (hss, bry) ~Tis( hss, ben) 1 (hss, bad) T (6.25 OBS O15 AIRS O.125, ons, Zu) Result -9 | = if T (bye) has a finite mean, then does the ane step basterior Probably. Peoaf ERE us concider the case, Ohin cot have gol a success the wwe need to prove that M Ths, ¢ is finite. oe Given Min(iyey "2 finde “0 (hs, t44) © = Tn Chs, t+ 4) 7 Tay ChE) Dag, = eee } nyo " & Ne Gy FE , ok,- 74 Fy Ce) Pn nyo St TM My [4% i <4] nyo : < K Z bal) Ty ay Chart) > nyo Sy 3 WC. ‘ . y Chy pe ) kK AT (h/t) Lek us consider a case Oheve we have qot a failuve, Spun Ale need ko prove Haak Mong dae) i finite Batata) 7 L* THlhG HY) se ? nyo a Th Gy) Un 0° eS nzZ0 6D Ted Lyo es, LZ A Thy Uyt) dm nyo . a ZT, Ct) ( As 4st] nd eer = taney — Hence froved (Ard.i : byt) After feabing Mone ease aud stavhiag with Ty Op ‘ _ ind We itl try fo find out expected Gains to 5 probability of SUCeRSS Expected gain = & (Mt) —¢ 2 4 > HOOPS Cy0 If x Carly) Tercis ao point in Festing. nase fg —ae- moc ChCP t SSO «ited Masiaramexpecked=ffanr Maximum eapected qain = max fo, «Catys})~0} =4,Cr(htJ) define, 4,(T (yt) =0 Ip, (TGA) = mort 9 “cm (nt)) (gy (lms 49) J TC COE) a, Creag, tay) 0= 4, (MaH)s Geta) & hela ae 4 KOGA) <3 SQuemey. a; (uty ee SSJresult 4 re If Oat) ts Finite Mean of T(t} then AM) S Ayr Br KYO ts 4 (wt) i, non Accveasing » there ave Some caovergtig upper band d (Tt) = lim Ay Coy T(hyt)) exists > maximum axpecked = Cumutahive wet gai, if we start with. Tht) n Proof = cosidey K=O, q, (taal) -0.¢ May (a,b) Let us Assume thal, I(T lht)) < Ay (4,8) for some K>1 aud ral W(u4) which har finite mean. Ta chow, Gea Ung) < ATC) trom vesuht 2 QuOn Urs tan) < Aw Cbs t4Y) Au (MD) SM "(ng bt) * Ayegy (CH A) = Yoo Sa, ca x(n as)) (24 9 Cts. 0)) ay 4 (1= 4 (CGA) Sa (ACHE AY) $ < a(ntyay)4 A (TCH) 4 CT Ch tN) ) le aun OnyH)) I CECH F849)2 ei oe >< ACM nt)) 4 (met) $A (nts: 4 (= Ur (nt) Aa (nt, OY) now, &Cr(h H) oe ht - Che) §C ee [te «tqauyys BOO Se By mA) DV tr | ” ney, i : W Thy (hs, t44) i At (ns ta) = y i Toa Ct) Pay | : n a | nyo MCnH Pe tZo | © Sy Rina (Ht) Bray D0 A (1(h,4) A (ID) Mag tay) 7 zn Tay Crt) Pray aati An (oh : = nus (MH, by) son nba ty ua I= X(n(h))) i a(n A)l Patan) > 2m Ty hg, : "0 » ¥| ee me hy {Wty < ae Te Gyt) bot zo Mnay CoA) Pay + ENT GH Vy n>0 G = Thy Chit) by + Th Chad Po Ty hd) b, £20 UH bs tor - +O Cnt) vy > 2M CG yt > - £2 1 Math ns &- eae art) < Arua Ve? Ven Gye HCE, Bp) = g CT) ra q Cathe t9)] 2) iuele, 245 (n(n) aeePensrtoe - te a cat) q (ong +40}} ~Sipe Rule. Tf ue short with ThA) 4 a Priov dishibuhon of remaining bugs having a finite ON thy: toh will continue teshing if q rw) >9 and stop tetis,, Jil CHt)) 0 fv song Ki Ta (hsf, £42) Ty (hs, faa, SG es D Ti(hs te) UZ. cz Tha (4b Pass n- CE Me Gt) P:) (2 iC hs 4) 44) 0 0 =m Tha Oh Pon Ue Tr, (nfs, b4d Tha ( hf try P, +} (Zn 669n9) Tay Cyt) Pasi nay CZ mi (mf e0 bi) ( = Tes) Li) oe ; i "nas (ht) Par Vinalfar Lel’s assume, Ta Chef, baa) = Th (hfs be2) t % Tyme mand) Pay Im ky That bk) ® te. 4, on CAP ; A a, * constant Veer ee cs Wa : Ae Ae G ae , 4, sae n In? % As Wn increases 7 Ay Arerrases Geometrically. — > Likelihood tahe ordering of dishibubons. This is used 2 ae tao'dishisations W! (hy by) Sa (ha, ts) Suppose for the prior 1 (hah), wt have decided to continue heshng > 2 4 (7 Oih)) >o > (lt) 21 (ha)ic \ A Wh by: {Mh (w,4) Th Cyt), ee ae Ta aed avd (ih) = {ty yh) albsh)- j sk 2 We (ha ta) od \Q0 Wwe wilt say at! Ch, ty) ts preferred jo T (he,2,) ly LR en ae vetahion, W' Chit) yoTl (ha, be) mn” likelth voc raho sen gt, ‘ th Ta’ oyh) Wn (hs) ) wherever, Th, hy #1) 2 Thy (hob) =y those canbe er cluded. Cael R [Reutt4] 5 1 Chto > Tr lba hy) thay Ty (yh a n a eee wv kyn & is nan decreasing mon (Peowf| a fs per definihon, ey t tf . Ty Chit) ‘ Ty Chae) : TH thy Ny Chats) Mt, Cha, fy) e q; nyt)Be TES gc ahh) Tr! Cag hd: Te th, ) oy Me Cri Ai) K 2 ne Gh) i ae Thar Ch, hy) See a 1 2 Te Cmte)» May (hayb) v=0 lim Kea te: lim Tk Chih) some STG Cy) there ise No, such thak Hf K >No Th Cry 4) Th Cha fy) ” Without loss of generality, Wt can assume that No Ig the smallest integer for abich i€ KYNe, ' TM (hah) . Ti Lhyls): ( y > iP Ween CO ay, 7 Ti Chih) if KN, thay e Zi Mw Chyt) % ZA Cryb) ee ; Zk FS if ASN we Kno 2 omeGi ti) S Th hy h) 0 N70 subtinchng ty’ (hy b) <1 Gah) HOA) YF That hehy at n> 4 f sub tracking TM Cat) € < T, Chayby ; 5 | ZW Oi) 218, moby continuing upko No, : z fy hy FD % Ze Tal hy ty f af n/Coyty > Thy), thon, Ti’ Ch) >” 3.7m (byt Oe ey| ((hz,t4)) | / ‘ a : Proof] zh Tw (wh) > x Fo Ts Chay) 3 PRAM HA) og (th thy ta Final motivation. 5 it Ww Ch, n) ow The by) ACM lyn) > 4 (ulry 43) TRA ba ay aay [Pesfl> cag, boy = Mn (ht) Gy ee Width) Ve Oo LR THO) > ThE bey fiy Chttas) Fy 7 Th hb k.re “ non) 8 fap "5 dacreasing Sequence - A von creasing seqaence In 4 Tn (4b) : Tafa : LR TOS Th beg). tk Result — F] > If 7 Chi, 7] > voy? Han oA Ghis, Ea} _ WT (has, bs + and, eee Be a (nef, Est) rm 2 Ty (his, brat) pate Ka Tn (his bat) 2 Tne Cah) Pray Ky 4 ky fn Cys, hat) r 1, é Thay Ch bv) Ty Chis hay ea Magy Cha hy - hs WC, a) ‘a T (hy hy) sy The + af a) ps won decneasing seq, Tay (hat) fin Chis, bat Por ae 5 Was bal)Foon om Chis tiaty 7 W (bys, #49) ; f / fs agen, Thy Chit BAL) fin (hy ti) C—O : i (hy 454 Hh ChS Bal) a hyn) / Uf ; fe, Te wh) > TH Cna Aa) Lf, toon TH Unt bat) SPY bof bah Sister a ae [Rea ts arises) 4, (906 [esata] ~ Hf Ww Ay) ¥ T Chg ty) Hom - ACG) > 4 (7 G4) fof bin By or! Gt) K-70 * Thic suffrcent ty prof Hot, Gy, (Wi Cyt) % Ay CHA) > ¥ 4,9 for Ket, 4, (ay! (hy) = tmar fo, A(t) -3 4, (1 (ah) > max fo, « (gh ))-4 | eonnwatt S SC TIhb)) > % (HUs,4))rr Lek, us assume thot, 4 Gal thy 1) % 480m, 4) for some Ky¢ Grom vel 7 OS ty) 9, Ce Chas bia) aud. 4, (n'chs bay)) S Gi (wag § tea) 7 xa (7 (ha,As)) : = mae fe (meh te) (14 4, (a (hs t,00)] al & (mlhaj2))) dn (n (haf, fei} © Alm (mb) [14 4 Caos, 6 41))] 4 (eeckrcha ay) Crh Gl < (1 (hy) [14 4, CW Cys, hay] 1 CALM da (1 cenf, a eg, (a (buf, hat) + & (lo, be) ft 9 Or! tas 9) ~ $C tag, fall San he eteult 23Ye og 4. CH (aE ne0) a x (11/ (ut J) [t+ 3x (Crys 41) ~ dM Gng ea | 4 Ai (Wah)) £ Baio MCh, by) eee ee Proof of raull 8 e | | | | V+ Ay UT Chsp tag) - he (H (ng Fad) 90 ¥kyo fay, Ker g, rlny, ta)) = mar $a a % (the) -< 4, (THO J = max 5 0, x(n 5) 3} Tay (h, 4) Bat) ae tee Ge seeps Thy (h4,be)= SR My (nf, HY) = a i Ty (nt) ! Raa Ty (hs, 1) Bue Ee fin (wt) ) Maa