2007-01-16 LiTH RUT - developmet hadbook 1.3 The Spiral Model v 4.0 Fredrik Herbertsso ABSTRACT The idea behid the spiral model is to do system developmet icremetally while usig aother developmet model, such as the waterfall model for each developmet step. This ca help mitigate the risks by ot requirig the etire system to be defied at the start of the developmet. Istead, the most importat parts are defied ad implemeted first. Usig the experiece gaied from this you the go back to defie ad implemet other parts of the system ad so o. To make use of the spiral model i a PUM project, a simplified versio should be used. Some suggestios o how the spiral model ca be adjusted to fit the PUM course are preseted i this documet. RUT - developmet hadbook 1.3 The Spiral Model v 4.0 1
Field of applicatio 1 Field of applicatio The spiral model where first proposed of Barry Bohem i a article from 1988 (Bohem, 1988). The spiral model is a developmet model that combies the iterative ature of prototypig with the cotrol ad systematic developmet of the sequetial model. The spiral model is well suited for developmet of ew products as well as performig improvemets o existig products. It s best used with medium to large projects with a log time frame, but it ca be applied to smaller projects. I his article (Boehm, 1988) Bohem writes The major distiguishig feature of the spiral model is that it create a risk-drive approach to the software process rather tha a primarily documet-drive approach or code-drive approaches. By its geeral ature ad risk-drive approach the spiral model ca coverge agaist other mai process models give how the risks are divided i the project. Some examples whe this happes is give i (va Vliet, 2005) Establishig the system requiremets is hard ad great risk is associated with this step. This ca be solved by goig a few cycles i the spiral model, which becomes aalogous to a developmet model based o prototype developmet ad evaluatio. The requiremets o the system are well defied. The spiral model ca be used to esure a robust ad well documeted system by lettig the cycles i the spiral model imitate the steps of the waterfall model. The spiral model is clearly very flexible ad ca be applied to almost ay project. 2 Prerequisites Accordig to (Boehm, 1988), the coditios that must be met if the spiral model is to be used are: The problem to be solved ca be divided ito several subproblems. The various subproblems have differet levels of risk associated with them. It is possible to implemet the system by first implemetig the parts that have high levels of risk associated with them, ad the move o to parts with lower risk levels. To aquire this data a iitail study of the domai could be eeded, this ca be performed as a seperate step before usig the sprial model or it ca be icorperated by usig a itial cycle. 3 Realizatio All iformatio preseted i this chapter has bee collected from (Wohli, 2006) ad (Pfleeger, 2006) if ot stated otherwise. The model described here is a later versio of the origial model proposed of Bohem i (Bohem, 1988). 3.1 Overview The spiral model is a variat of iterative developmet. It tries to combie risk maagemet with iterative developmet. The spiral model is built aroud these keywords (Wohli, 2006) : Budget 2 RUT - developmet hadbook 1.3 The Spiral Model v 4.0
Realizatio Detrmie Goals Alteratives Costraits Risk aalysis Develop ad test The basic idea of the spiral model is to maage ad miimize risks by repeated use of risk aalysis, prototypes ad evaluatio. The developmet starts at the ceter ad proceeds clockwise through the spiral, progressively buildig a more complete system. The spiral is divided ito four quadrats each represetig a geeral cocept of what is to be doe; each quadrat cotais a certai amout of steps to be performed depedig o the complexity of the project. Whe each step has bee performed oce, oe cycle of the model is completed. A geeric project usually cotais 4 cycles plus the evetual maiteace cycles, but a project ca cotai ay amout of cycles. Figure 3. A typical spiral model with 4 cycles The result of a cycle i the spiral does ot have to be a fuctioal system. Istead it ca be a requiremets specificatio, desig documet, low-fi prototype or some improvemets o a pre-existig product. The ed result of each cycle is selected depedig o the size ad type of project. Geerally, each cosecutive cycle delivers somethig that is closer to the fial product. The same cycle is repeated for each level of refiemet. This RUT - developmet hadbook 1.3 The Spiral Model v 4.0 3
Realizatio applies to the etire developmet process, from the first sketches right dow to the implemetatio of each compoet. As opposed to classical developmet models (e.g. the waterfall model) the spiral does t ecessarily ed upo delivery of the completed system. Istead, the spiral model ca be applied to the etire life cycle of a product. 3.2 Detailed descriptio As previously metioed, each cycle is divided ito four quadrats. For each quadrat a umber of steps represetig various tasks are specified. The ature of these tasks depeds o the size ad complexity of the project. I a small project the tasks are geerally fewer ad less formal compared to a large scale project. The four quadrats i the spiral model gives the workflow of the sprial model. Due to the geeral ature of the spiral model, it s hard to give a absolute work descriptio, but the model gives good guidlies how to work. Quadrat 1: Determie goals, alteratives, costraits This quadrats tasks are: Budget Goals Alteratives Costraits The tasks i the first quadrat aims to decide the cycles budget, determie the differet goals or objectives to be implemeted i this portio of the project, idetify the differet alteratives to implemet the goals ad determie the costraits o the cycle ad idetify differet costraits imposed by the alteratives. Quadrat 2: Evaluate alteratives ad risks This quadrats task are: Evaluate the goals ad costraits Risk aalysis Prototypig I the secod quadrat the differet alteratives are evaluated i relative to the goals ad costraits. If areas of ucertaities are idetified durig the evaluatio, a risk aalysis is performed to aalyse the correspodig risks. Whe these risks have bee idetified, cost effective solutios to the potetial problems have to be formulated, this is doe by prototypig. Prototypig should be cosidered i geeral terms ad does t ecessarily mea that differet versios is made for each alterative istead it meas that sufficiet iformatio is collected to make a throughout ivestigatio e.g. cosult a expert, questioaires or prototypig. Quadrat 3: Develop ad test This quadrats tasks are: Implemet Validate/Test I the third quadrat actual implemetatio is doe. If the domiatig risks reside i e.g. user iterface or ay other area where further prototypig ca solve the risk, the task per- 4 RUT - developmet hadbook 1.3 The Spiral Model v 4.0
Results formed here could be pla ad develop a more detailed prototype or make a evolutioary developmet of the product. If these risks are cosidered solved or low risk the model follows the basic waterfall model e.g. requiremets specificatio, software desig etc. is produced, as ca be see i figure 3, see RUT 1.1 Waterfall model for further iformatio about the waterfall model. Quadrat 4: Pla This quadrats tasks are: Pla ext phase I the fourth quadrat of the model all the products produced durig the cycle are evaluated. The evaluatio is the foudatio for the plaig for the ext cycle this icludes settig milestoes ad deadlies, dividig resources ad ay other plaig withi the project. This steps always results i a revised versio of the project pla. The cycle is completed with a review doe by everyoe affected (developers, customers ad maybe users) by the product. The reviews purpose is to verify that all parties have agreed o the ext cycles course of actios. 4Results 4.1 Products The spiral model is used i system developmet projects, ad the fiished system is the mai product. The model ca also be used whe maitaiig or improvig a pre-existig system i which case the revised system is the mai product. The sprial model also produces a umber of documets: Cocept of operatio Requirmet specificatio Developmet pla Software desig pla Itegratio ad test pla User maual The ame ad amout of documets produced vary from project to project but these documets should be cosidered essetial ad therefore by part of the spiral models products. 4.2 Product templates The article by Boehm (Boehm, 1988) cotais suggestios for a possible product template where iformatio regardig the work performed i each spiral cycle ca be etered. These templates where origially used oly for presetatios, but later became a stadard part of all projects. Table 1 cotais oe such possible template. Goal Limitatios Tabell 1: RUT - developmet hadbook 1.3 The Spiral Model v 4.0 5
Templates ad forms Tabell 1: Optios Risks Suggested solutios to risks Results of risk maagemet Plas for the ext phase Commitmet 5 Templates ad forms Nearly all RUT:s are applicable to the spiral model. Exactly which RUT:s that are relevat to a particular project depeds o that project, but some geeral suggestios ca be foud below. RUT 3.1 Developmet of a Project Pla RUT 3.8 Risk aalysis RUT 4.1 Customer Relatios RUT 4.2 Domai aalysis RUT 7.1 Choice of desig method RUT 10.2 Plaig uit testig RUT 12.1 Choice of test strategy (val av teststrategi) RUT 13.1 How to fid system test cases RUT 14.1 Acceptace test Ay RUT pertaiig to implemetatio, sectio 9. Each documet has a set of forms ad templates that ca be used durig the developmet process. 6 Verificatio of results Verificatio of the results depeds etirely o the methods used durig the developmet. 7 Examples with explaatios The followig example is a scaled dow versio of the example give i (Boehm, 1988). The article shows three cycles while this example oly has two. The project i the example is the developmet of a software system called TRW Software Productivity System (TRW-SPS), a itegrated software egieerig eviromet for the america military. 6 RUT - developmet hadbook 1.3 The Spiral Model v 4.0
Examples with explaatios The developmet of TRW-SPS started with a eed for icreased productivity i the military, ad a hypothesis that software egieerig was a good place to start. Because of this, the project started with a mii-cycle with the purpose of evaluatig the feasibility of icreasig the software productivity at a reasoable cost. Cycle 0: Usability study As ca be see i table 2, the goals ad limitatios i this cycle are very high level, geeral statemets like sigificat gai ad reasoable cost. Table 2: Goal Limitatios Optios Risks Suggested solutios to risks Results of risk maagemet Plas for ext phase Commitmet Icrease software productivity sigificatly Reasoable cost. Govermet cotracts, high tech, huma orieted, security. Maagemet: Project orgaizatio, policies, plaig, cotrol. Staff: Maagemet, simulatios, traiig. Techology: Tools, workstatios, methods, reusability. Facilities: Offices, commuicatios. Great opportuities for icreased productivity may be missed. Improvemets may coflict with limitatios. Iteral ivestigatios Aalyze the cost model Aalyze exceptioal projects Literature searches Some solutios ca ot be realized. A time-sharig system: security. A mixture of alteratives may give great gais A factor of two i five years Additioal ivestigatio eeded to determie optimum mix. Six people to work for two moths More advaced ivestigatio ad aalysis Iteral, exteral ecoomy Develop work ad ecoomical guidelies. Fiace ext phase Some of the alteratives cosidered uder Techology might lead to the developmet of software, but the developers decided agaist a software-based solutio as a hardware solutio held great promise. Oe possible risk metioed i table 2 is the possibility of makig major ivestmets oly to fid it impossible to reach the ehaced productivity goals. To try ad miimize this risk, a umber of ivestigatios ad aalysis were performed. This icluded iterviews with software developers och maagers, a aalysis of productivity ehacig fac- RUT - developmet hadbook 1.3 The Spiral Model v 4.0 7
Examples with explaatios tors with the Cocomo-model ad a aalysis of previous productivity ehacig TRW projects. Results of the risk aalysis poited to the possibility of icreased productivity at a reasoable cost through the use of a sigle maiframe ad time-sharig. This solutio was ot compatible with TRW rules regardig security however. There was a requiremet for multiple security levels for classified projects. As you ca see, some suggestios could be elimiated early o, at this high abstractio level. The pla for the ext cycle called for aother twelve ma-moths to be ivested i the projects, compared to oly two durig the first. Cycle 1: Behavioural descriptio Table 3 cotais a overview of the work doe i cycle 1, just as table 2 did for cycle 0. The mai differeces betwee the cycles are as follows: The ivestmet level is higher i cycle 1 (twelve ma-moths compared to two). Goals ad limitatios are more clearly defied ( double software productivity i five years at a cost of $10000 per perso versus sigificat gais at a reasoable cost ). More limitatios itroduced, e.g. preferece for TRW products (specifically a LAN developed by TRW). Alteratives more detailed ( Requiremet tools SREM, PSL/PSA or SADT, etc.) Potetial risks more clearly defied ( TRW LAN price/performace for less tha $10000 per perso vs. improvemets may coflict with reasoable costs ). More effort was put ito maagig the idetified risks (this icluded performace measuremets o a prototype of TRW LAN which was i developmet for aother project). The result was a fairly specific behaviour documet, cotaiig custom made offices for software developmet ad persoal termials coected to VAX miicomputers via TRW LAN. Some choices were postpoed util the ext cycle, e.g. choosig the operatig system ad tools to be used. The life cycle pla ad the pla for the ext cycle cocluded that the project should be divided ito several idepedet spirals to improve the project cotrol, ad that developmet of software utilities ad a first versio of a Software Developmet Eviromet should begi. The commitmet step meat a bit more tha simply decidig the pla for the ext cycle. It was decided to test the software by usig a 100 perso software project to develop a testbed ad a eviromet focusig o the eeds of the testbed project. The pla recommeded the developmet of a prototype eviromet, but also the use of requiremets specificatios ad desig specificatios i a risk-drive way. Table 3: Goal Limitatios Double software productivity withi five years. Ivestmet of $10000 per perso. Govermet cotracts, high tech, huma orieted, security Prefer TRW products over others. 8 RUT - developmet hadbook 1.3 The Spiral Model v 4.0
Solutios to commo problems Table 3: Optios Risks Suggested solutios to risks Results of risk maagemet Plas for ext phase Commitmet Office: Private/modular/... Commuicatios: LAN/star/cocetrator/... Termials: Private/shared; smart/dumb Tools: SREM/PSL-PSA/...; PDL/SADT/... CPU: IBM/DEC/CDC/... May miss thigs with a great potetial. TRW LAN price/performace Workstatio costs Extesive third party ivestigatio, visits. TRW LAN performace measuremets. Ivestigate price of workstatios. Operatioal cocepts: Private offices, TRW LAN, persoal termials, VAX Begi with dumb termials; experimet with smart workstatios. Postpoe selectio of OS. Split work ito Software Developmet Eviromet (SDE), utilities, ehaced project maagemet Develop a first release of SDE Desig may cost: 15-people team for oe year. Pla for use of third-party. Develop prototype of SDE Get a ew project to use SDE Make SDE support the project Create represetative cotrol group 8 Solutios to commo problems The author of this documet feels that he have to little experiece usig the sprial model to give ay commo problems ad their solutio. 9 Adaptatio to the PUM course As previously metioed, the spiral model ca be viewed as a aid to help with selectig the right model for every phase of a project. The spiral model is therefore suitable i a wide rage of areas. I order to use the full potetial of the model, it should be applied to a relatively large project that lasts for a exteded period of time. See sectio 7 for a example of this. For a smaller projects, such as a PUM project, the model eeds to be adjusted as there is very little time available for aalysis ad re-plaig. I a small project the spiral model may ot reach it s full potetial ad may loose some of it s advatages. I order to take advatage of the model i a PUM project it is better to use a simplified model that uses some of the basic ideas of the spiral model. The advatages this brigs comes from beig able to streamlie the iteratios performed i e.g. the RUT - developmet hadbook 1.3 The Spiral Model v 4.0 9
Measurig the process waterfall model. The repeated evaluatios ad re-plaig advocated i the spiral model ca improve the result of a PUM project cosiderably. The deadlies for the various documets i the course seem to be suited to a sequetial method divided ito phases, such as the waterfall model. However, the spiral model ca still be used successfully i the PUM course! Oe possibility is to use the spiral model for each phase of the developmet process (desig, implemetatio etc.) If the project group iteds to improve or exted a existig product, produce several differet prototypes or if the project ca be divided ito idepedet subprojects, the spiral model may very well be the optio available. 9.1 What to iclude? The followig parts of the spiral model should be preserved i order to use it effectively i a smaller project: Formulate the goals of the cycle, as well as possible solutios ad limitatios. Evaluate the differet possible solutios ad idetify possible risks. Perform the work associated with the cycle ad review or test it. Evaluate the work doe ad pla for the ext cycle. 9.2 Phase plaig I order to create a simplified spiral model oe should carefully cosider how may cycles to iclude. A pla outliig which phases of the PUM project to iclude i each cycle must be created. Suggested adaptatios To adapt the spiral model to the PUM course, three cycles should be used i the spiral. The phases of the PUM project should be mapped to the cycles as follows: Cycle 1: Pre-study ad defiitio. Cycle 2: Architecture ad Desig Cycle 3: Implemetatio ad test. The exact cotet of each phase, ad which activities from the spiral model described earlier to iclude depeds o the properties of the idividual project. This detailed plaig must be doe early i the project ad should be icluded i the project pla. The after study is ot icluded i the spiral because it is a evaluatio doe separately from the project itself. 10 Measurig the process Fidig some absolute measure of how well a system developmet model works is ot a easy task. The mai problem is that the process ca be applied to such a wide rage of differet projects. The goal is therefore to fid some project idepedet variable to measure. 10 RUT - developmet hadbook 1.3 The Spiral Model v 4.0
Measurig the process 10.1 Resource measuremet The resource measuremet is meat to show mow much resources the method cosumes. Oe suggestio is to measure how much resources are spet o admiistratio. 10.2 Product measuremet Oe way to measure how well the spiral model is workig is to estimate how much time is spet o uecessary work. Here, uecessary meas ay time spet o tasks that later tur out to be dead eds. The percetage of time spet o these tasks is a idicatio of how fast ad effectively the process idetifies ad corrects these dead eds. 10.3 Forms for data collectio I order to estimate the effectiveess of the process, the time spet o various tasks must be carefully oted ad examied. This ca be doe i the followig way: the total time spet is logged as the work progresses. Whe a dead ed is discovered, the total amout of time spet o that particular track is estimated. Exactly how the time is logged depeds o the particular project, but it is importat that the total ad dead-ed time is calculated cosistetly withi the project. Sample form ad usage Table 4: Name of project ad perso Week Admiistratio Total time Wasted time 10 2 8 6 11 0 14 4 12 3 18 3 13 1 20 2 Sum: 6 60 15 Wasted percetage = 15 / 60 = 25% Admiistrative percetage = 6/ 60 = 10% RUT - developmet hadbook 1.3 The Spiral Model v 4.0 11
Documet history 11 Documet history Table 5: Versio Date Descriptio Author 1.0 95-02-16 Documet created. Joh Olsso 2.0 97-02-14 The ew spiral model added Measurig the process added Forms ad templates added. Ulf Blad 2.1 97-06-03 Adaptatios to the PUM course added. Ulf Blad 2.2 98-06-11 Laguage ad structural chages. Ja Carlso 2.3 00-06-13 Chapterl 3 completely rewritte, additios to chapter 4, 5 ad 9 ad a umber of mior chages. Fredrik Motelius 3.0 03-01-07 Traslatio ito eglish Erik Gustavsso 3.1 06-12-27 Chaged chapter 1 ad 2. Fredrik Herbertsso 3.2 07-01-08 Completely rewritte chapter 3. Added exteded chapter 4 with more produts. Fredrik Herbertsso 3.3 07-01-10 Deleted the ew sprial model i chapter 8. Fredrik Herbertsso 4.0 07-01-14 Updated chapter 5, made mior chages i chapter 9 ad updated the refereses. Fredrik Herbertsso 12 Suggested improvemets The example i chapter 7 must be rewritte to be more clear or chaged for a brad ew example. 13 Refereces 13.1 Method descriptios Boehm, Barry W. (1988) "A Spiral Model of Software Developmet ad Ehacemet" Computer 1988:5, 61-72 Pressma, Roger S. (2005) "Software Egieerig - A practitioer?s Approach, 6th Editio" McGraw-Hill Compaies ISBN: 007 301 933-X Vliet, Has va. (2000) "Software Egieerig? Priciple ad Practice, 2d Editio" Joh Wiley & Sos ISBN: 047 197 508-7 12 RUT - developmet hadbook 1.3 The Spiral Model v 4.0
Refereces Wohli, Claes (2005) "Itroduktio till programvaruutvecklig, 1st Editio" Studetlitteratur, ISBN: 914 402 861-X Pfleeger, Shari L., Atlee, Joae M. (2006) "Software Egieerig -Theory ad Pratice, 3th Editio", ISBN: 013 198 461-6 13.2 Evaluatio of the method Wolff, J. Gerard (1989) "The Maagemet of Risk i System Developmet: Project SP ad the New Spiral Model." Software egieerig Joural 1989:4, 134-142. Weckma, J., Colvi, T., Gaskis, R.J., Mackulak, G.T. (1999) "Applicatio of simulatio ad the Boehm spiral model to 300-mm logistics system risk reductio" Simulatio Coferece Proceedigs 1999:1, 912-917 13.3 Iteral commets No iteral commets. RUT - developmet hadbook 1.3 The Spiral Model v 4.0 13
Refereces 14 RUT - developmet hadbook 1.3 The Spiral Model v 4.0