TotalCostofPreservation(TCP) Page1of20
TotalCostofPreservation(TCP):
CostModelingforSustainableServices
Rev.0.1620120409
StephenAbrams
PatriciaCruse
JohnKunze
UCCurationCenter,CaliforniaDigitalLibrary
UniversityofCalifornia,Officeof thePresident
MichaelMundrane
OfficeoftheDeputyCIO
UniversityofCalifornia,Berkeley
1 Introduction
Informationtechnologyandresourcesarethoroughlyintegratedwith,andindispensableto,
today’sInternetandwebbasedculture,commerce,science,education,andentertainment.The
digitalassetsunderpinningtheseactivities,however,areinherentlyfragilewithrespecttoever
increasingdisruptivete chnologicalchange.Withouteffective,affordable,andproactivecuration
management,today’sdigitalassets
willnotremainviableandusefulinthefuture.Inorderto
achieveubiquityoflongtermpreservationeffortsandresults,thefulleconomiccostsof
preservationactivitiesovertime,thetotalcost ofpreservation(TCP),mustbewellunderstood.
TCPanalysiscanbeappliedusefullyinthedevelopmentof
twospecificpricemodelsfor
preservationservicepricing:
Payasyougo
Paidup
Thepayasyougomodelisappropriateforsituationswhereareliableandpredictableannual
incomestreamisavailabletotheclientpurchasingpreservationservices.Whenthisisnotthe
case,forexample,
withorganizationsfacingirregularorboomorbustbudgetarycyclesorfor
grantfunded,fixedtermresearchprojects,thepaidupmodelmaybemoreattractive;indeed,in
manycircumstancesitmaybetheonlyrealisticoption.
2 Totalcostofpreservation
Longtermdigitalpreservationisacomplexactivity,involvingsophisticatedtechnical
infrastructureaswellassignificanthumancompetencies,analysis,anddecisionmaking.Given
thedifficultiesinaccounti ngforthemyriadaspectsofpreservationactivity,andofprojectingthe
UCCurationCenter
TotalCostofPreservation(TCP) Page2of20
costofthoseactivitiesintotheindefinitefuture,any analysisofTCPmustrelyonanumberof
fundamentalassumptions.(SeeAppendixAforasummaryofpriorworkonpreservationcost
analysis.)
2.1 Assumptions
TheTCPanalysisassumesthefollowing:
Onlythecostspertainingtopreservationserviceprovidersareconsidered.Intermsofthe
OAISreferencemodel[OAIS],thesearethecostsassociatedwithanArchive,“an
organization[ofpeopleandsystems]…thathasacceptedtheresponsibilitytopreserve
information.”Inparticular ,thepreservation
costsassociatedwiththelocalactivitiesof
OAISProducers(e.g.,contentcreationoracquisition, reformatting,packaging,
submission,etc.)areconsideredoutofscope.Ontheotherhand,thecostofsupporting
aProducerinmakinguseofArchivefunctionsisinscope.
Individualcostcomponentsofthemod elcanbe
categorizedunambiguouslyaseither
fixed,whichareincurredregardlessoflevelofuse,ormarginal,whichscale
proportionallywithuse.
Individualcostcomponentscanbecategorizedunambiguouslyaseitheronetimeor
recurring.Onetimecostscanbeannualizedovertheeffectivelifespanoftheactivityor
component.
Thevaluesdefinedforindividualcostcomponentsrepresentnominalcosts,thatis,the
costsdefinedforgenericinstancesofactivitiesorsystemcapacities.Thisisreasonable
underafurtherassumptionofapolicyofuniformityofpreservationeffortratherthan
outcome.Forexample,onthebasisofform,structure,
accompanyingmetadata,etc.,
somedigitalcontentmaybemore inhe rentlyamenabletopreservationcareandwill
naturallyreceiveahigherlevelofpreservationserviceandoutcome.
Actionsperformedonpreservationcontent,e.g.,characterization,fixity,tra nsformation,
etc.,aresubstantiallyautomated.Thus,themaincostassociatedwithanactionisin
the
acquisitionoranalysis,design,anddevelopmentanddeploymentoftheimplementing
software,whichisindependentofthenumberofobjectsagainstwhichtheactionis
performed.
Thesizeandscopeofthecontenttobesupportedisknown,andthecostofprovidingfor
itspreservationis
established,atthebeginningofthetimeperiodunderconsideration.
However,underthepayasyougopricemodelthesecostsarebilledforattheendofthe
timeperiodwhileunderthepaidupmodeltheyarebilledforatthebeginningofthe
period.
Thepreservationserviceprovidercancarryforwardbudgetarysurplusesacrossfiscalyear
boundaries.Furthermore,thesesurplusfundscanbeinvestedatmarketrates,withthe
UCCurationCenter
TotalCostofPreservation(TCP) Page3of20
earnedinterestfurthercontributingtothesurplus.
Themodelsareultimatelyrevenueneutral,thatis,thereshouldbenosurplus(ordeficit)
fundsremainingattheendoftheperiodunderconsideration.
Valuescanbedetermined
forvariousannualadjustmentfactors,suchasinflation,
investmentrateofreturn,costoflivingadjustments(COLA),meritpayraises,and
changestotheperunitcostofpreservation,which,althoughheldconstantoverthefull
periodofTCPconsideration,areneverthelessreliablypredictiveoflongtermeconomic,
technological,and
organizationaltrends.
Thelastassumption,althoughthemainstayofthestandardeconomicforecastingtechniqueof
discountedcashflow(DCF)analysis[Damodaran],canbeproblematicoverextendedtime
periods.ConcernsovertheefficacyofDCFbasedmodelingareaddressedinSection§3.
2.2 TCPanalysis
TheTCPanalysisencompassesthefulleconomiccostsassociatedwiththelongterm preservation
ofdigitalassets,althoughtheresultingpricemodelscanbeeasilycustomizedtodealonlywith
varioussubsetsofthosecostsasamatteroflocalpolicy.Theindividualcostcomponents
incorporatedintheanalysisarereflective
ofthefollowing11highlevelcategoriesabstracted
fromtheenvironmentinwhichpreservationactivitiesoccur(seeTable1):
Preservationactivitiesareembodiedinanarchival
1.
System;composedofvarious
2.
Servicessupportingnecessaryanddesirablefunctions;runningon
 3.
Servers;designed,deployed,maintained,enhanced,andutilizedby
 4.
Staff;insupportofcontent
 5.
Producers;whouse
6.
Workflowstosubmitinstancesof
7.
ContentTypes;whichoccupy
 8.
Storage;andaresubjecttoongoing
 9.
Monitoring;andperiodic
 10.
Interventions;allsubjecttoappropriatemanagerial
 11.Oversigh t.
Table1Preservationenvironment
EachoftheprecedingitalicizeditemsisacostcomponentoftheTCPanalysis.Tosimplifythat
UCCurationCenter
TotalCostofPreservation(TCP) Page4of20
analysis,theSystemcomponentencompassesthebaselinetechnicalinfrastructureand
environmentinwhichpreservationactivitiesareembodied,andinparticular,subsumesthe
ServicesthatsupportnecessarytechnicalfunctionsandtheServersonwhichthoseServicesare
deployed.Similarly,
Staffcostsaresubsumedbytheotherindividualhighlevelcostcompon ent
terms.
TheContentTypecomponentencompassesallanalysis,planning,andsoftwareacquisitionor
developmentspecifictosupportinggivenclassofdigitalcontent.
TheMonitoringcomponentencompassestechnologywatchforincipientobsolescenceandother
riskfactorsthatmayimpinge
onlongtermusability;routineproactivemonitoringoftheSystem
(anditsunderlyingServicesandServers)andStorageforavailabilityandresponsivity,ontheother
hand,aresubsumedundertheSystemandStoragecomponentsrespectively.
TheInterventioncomponentencompassesallnonperiodicorunpredictablepreservation
activitiessuchasformat
migrations,responsestorightschallenges, disasterrecoveryefforts,etc.,
storagemediarefreshandmigrationaresubsumedundertheStoragecomponent.
TheTCPanalysisrepresentsthetotalcostofpreservationtothepreservationserviceprovideras
follows:

··ℓ··
·· (1)
TC
P
TotalcostofpreservationforallProducers.
A
FixedcostofthebaselinearchivalSystem.
n
NumberofcontentProducers.
P
UnitcostofsupportingaProducer.
NumberofsubmissionWorkflows.
W
UnitcostofsupportingaWorkflow.
NumberofContentTypes.
C
UnitcostofsupportingaContentType.
k
NumberofunitsofpreservationStorage.
S
UnitcostofStorage.
j
NumberofpreservationMonitoring activities.
M
UnitcostofaMonitoring activity.
i
Numberofpreservation Interventions.
V
UnitcostofanIntervention.
O
FixedcostofadministrativeandmanagerialOversight.
Table2TCPnotation
UCCurationCenter
TotalCostofPreservation(TCP) Page5of20
ThebaselinearchivalSystemandadministrative/managerialOversightareconsideredfixedcosts,
thereforetheyarerepresentedinEquation(1)bysingleterms,
A
and
O
, respectively.Allother
componentsaremarginalcosts,eachre presentingbytwoterms:aunitcostandthenumberof
unitsallocatedorconsumed.
TheTCPanalysiscanbeusedasthebasisforpartialcostrecoverybyremovingorzeroingout
particularcostcomponentsthataresubsidizedasamatter
oflocalpolicy.Thespecificpayas
yougoandpaiduppricemodelsarebasedonaderivationofthepreservationcostsattributable
toagivencontentProducer.
2.3 Payasyougopricemodel
Thepayasyougopricemodelisbasedonanannualbillingcycle,
t
 0,1,2,….Certaincost
componentsSystem,Workflow,ContentType,Monitoring,Intervention,andOversightare
considered“commongoods”;thatis,theyareapplicableandbeneficialtoallPr oducersequally,
andassuch,areproperlyapportionedacrossallProducers.ThenumberofunitsofStorage,on
theotherhand,
isspecifictoasingleProducer.Theannualpayasyougopriceforagiven
Produceris:

·ℓ·
··

·
(2)
G
Payasyougoprice.
k
P
NumberofunitsofpreservationStorage attributabletoagivenProducer.
Table3Payasyougopricemodelnotation
Theoperationofthepayasyougomodelisillustratedinthecashflowdiagram[Neftci]inFigure
1.(Valuesabovethehorizontaltimeaxisrepresentincomeorprice;thosebelowtheaxis
representexpensesorcosts.)Aseries
ofequalcosts
G
areincurredattheendofeachservice
year.Toachieverevenueneutralitythesearematchedwithcorrespondingprices.
Income
G
G
G


t=0 1 2 3

Expense
G
G
G
Figure1Payasyougopricemodelcashflowdiagram
ThelongtermcosttoaProducerunderthepayasyougomodelisfoundbysummingtogether
theannualpayments
G
overaperiodof
T
years.
UCCurationCenter
TotalCostofPreservation(TCP) Page6of20
 (3)
· (4)
G
T
Cumulativepayasyougopriceover
T
years.
G
Annualpayasyougoprice.
T
Fixedtermof
T
years.
Table4Payasyougopricemodelnotation
Unfortunately,thiscumulativepriceincreaseslinearlyasafunctionoftimeand,mostalarmingly,
as
T
approaches“forever”(
T
),thetotalpricealsoapproaches infinity.
 (5)
However,recallingtheassumptionofanannualdecreaseintheaggregatecostofproviding
preservationservice,
d
,theseriesofannualcostsandpricesdecreasesuniformly,asillustra tedin
Figure2.
Income
1   ·  1  
·


t=0 1 2 3

Expense
1   ·  1  
·
Figure2Discountedpayasyougopricemodelcashflowdiagram
Thesizeoftheannualdecreaseisproportionalto1–
d
,whichcompoundsovertime.Again,the
longtermcostofisfoundbysummingtogethertheannualpaymentsover
T
years.
,
·
1
·1
·1

(6)
,
·
11
(7)
,
(8)
G
T
,
d
Cumulativediscountedpayasyougopriceover
T
years.
d
Annualpercentagerateofdecrease inaggregatepreservationservicecost
(asadecimal,forexample,0.1fora10%annualdecrease).
Table5Payasyougopricemodelnotation(continued)
Mathematically,Equations(7)and(8)arethesolutionsoffiniteandinfinitepowerseries
UCCurationCenter
TotalCostofPreservation(TCP) Page7of20
respectively,whicharesolvableonlywhentherateofdecreaseinaggregatepreservationservice
cost,1–
d
,islessthan1.Fordetailsofthederivationoftheseequations,seeAppendicesBandC.
Forthederivationofthevalueof
d
seeAppendixD.
2.3 Paiduppricemodel
Thepaiduppricemodelisbasedonaonetimeprice, paidatthebeginningoftheperiodunder
consideration,thatissufficienttofinanceallsubsequentpreservationactivitiesoverthatperiod.
Thisperiodcanbeeitherafixedterm
T
or“forever,”thatis,for
T
.Intherelevantcashflow
diagraminFigure3,theonetimepaymentFispaidbytheProduceratthebeginningofyear1(
t
 0).Thesubsequentincometermsaretheresultoftheinvestmentreturn
r
appliedtoeach
year’ssurplusfunds.Notethat
r
isthenominalannualpercentagerate(APR)ofinvestment
return.Theeffectiveannualrate,
e
,incorporatingtheeffectofmonthlycompounding,is
[Tucker]:

1
12

1
(9)
Thepayasyougocost
G
isscaledbyafactorof
e
/
r
toaccountforthefactthattheservice
providerincurscoststhroughouttheserv iceyearbutthesearenotbilledandpaidforuntilthe
endoftheyear.
Income
F
·
·

1
·
1
·
·

1
·

1
·
1

·

1

2
·



t=0 1 2 3

Expense
1   · 
1
·
Figure3Paiduppricemodelcashflowdiagram
Thepaidupprice
F
isderivedinAppendixE.
,,
·
·
1
1
 ·
·
1
1
·
·
1
1

·
·
1  

1  
(10)
,,
·
·
1
1
1
·
(11)
,,  ·
·
1

(12)
UCCurationCenter
TotalCostofPreservation(TCP) Page8of20
F
T
,
d
,
r
Paiduppricefor
T
years.
r
Thenominalannualpercentagerate(APR)ofinvestmentreturn(asa
decimal,forexample,0.02for2%.
e
Theeffectiveannualpercentagerateofinvestmentreturn(asadecimal).
Table6Paiduppricemodelnotation
Equations(11)and(12)arealsopowerseriesandaresolvableonlywhenthediscountfactoris
lessthantheeffectiveinvestmentretur n,1–
d
 1
e
.Fordetailsofthederivationofthese
equations,seeAppendixE.
ItisdesirableforaProducertobeabletoswitcheasilybetweenapayasyougoandpaidup
basis.Therelevantfactorinthistransitionisthe“coefficientofpermanence,”thatis,theone
timepremium,
ormultiplier,oftheannualpayasyougoprice,
G
,thatonemustincurinorderto
achievepaiduppermanence.
φ
·
(13)
φ
Coefficientofpermanence;theonetimemultiplierofthepayasyougoprice
necessarytoachievepaiduppermanence.
Table7Paidupmodelnotation(continued)
Thus,aProducercurrentlypaying
G
onanannualbasiscan“tradeup”toservicepermanencewith
aonetimepaymentof
·
·
.
Thepaidupmodelisdesignedtobeultimately revenueneutral,thatis,withnosurplus(or
deficit)fundsremainingattheendoftheterm.Thiscanbeillustratedinthefollowingexample
foratenyearterm(
T
=10),apayasyougoprice
G
of$650(for1TBofstorage),anominal
annualinvestmentrateofreturnof2%(
r
=0.02),andanannualdecreaseintheaggregatecost
ofpreservationserviceof10%(
d
=0.1).FromEquation(11),thisleadstoapaidupprice
F
of
$4,764.14.Incomeinyears1,2,3,...istheresultoftheinvestmentreturnonthesurplus
remainingattheendoftheprioryear.
Year(T) Income Expense Surplus
0 $4,764.14 $0.00 $4,764.14
1 $96.16 $655.99 $4,204.31
2 $84.86 $623.19 $3,665.98
3 $73.00 $592.03 $3,147.94
UCCurationCenter
TotalCostofPreservation(TCP) Page9of20
4 $
63.54 $562.43 $2,649.05
5 $53.47 $534.31 $2,168.21
6 $43.76 $507.59 $1,704.38
7 $34.40 $482.21 $1,256.57
8 $25.36 $458.10 $823.83
9 $16.63 $435.20 $405.26
10 $8.18 $ 413.44 $0.00
Table810yeartermpaiduppricemodelexample
Therelationshipbetweenthepayasyougoandpaidupmodels isillustratedinFigure4.The
cumulativeanddiscountedpayasyougopricesandthepaidupprice,
G
T
,
G
T
,
d
,
F
T
,
d
,
r
,
areshownasthered,orange,andgreencurves,
respectively.Thecum ulativepayasyougoand
paiduppricesfor“forever,”
G
,
d
and
F
,
d
,
r
,areshowninmagentaandblue,
respectively.
Figure4Relationshipbetweenpayasyougoandpaiduppricemodels
Thedifferencebetweenthecumulative(red)anddiscounted(orange)payasyougopricesis
accountedforbythediscountingfactor
d
,theassumedannualdecreaseinthecostofproviding
preservationservice.Thedifferencebetweenthefixedtermpayasyougo(orange)andpaidup
(green)prices,aswellasthe“forever”prices(magentaandblue),isaccountedforby
r
,the
assumedannualinvestmentrateofreturn.
Thecurveforthefixedtermpaiduppricecharacteristicallyapproachestheforeverprice
asymptoticallysolongastheinvestmentrateofreturnandannualdecreaseinaggregate
0
2000
4000
6000
8000
10000
12000
14000
16000
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Cost($)
Year(T)
Cumulativepayasyougo
price,G(T)
Discountedpayasyougo
price,G(T,d)
Discountedpayasyougo
price,G(,d)
Pai
d
upprice,F(T,d,r)
Pai
d
upprice,F(,d,r)
UCCurationCenter
TotalCostofPreservation(TCP) Page10of20
preservationservicecostaregreaterthan0(
r
 0 and
d
 0,oriftheaggregatecostincreases
(i.e.,
d
 0),whentheinvestmentgainisgreaterthantheservicecostincrease(
r
 |
d
|).
2.4 Rateofchangeofpreservationservice cost
Thevariable
d
inEquations(7),(8),(11),and(12)isacompositevalueorfunctionthatsumsthe
individualcomponentspecificchangesinquantityandunitcost,weightedbythecomponent’s
proportionatecontributiontotheoverallcost(seeAppendixE).
3 Predicativereliability
ManyoftheassumptionsbuiltintotheTCPmodelmaybeproble maticoversignificanttime
spans,particularlytheassumptionthatthevariousannualadjustmentfactorsinvestmentrate,
rateofdecreaseofunitcost,inflation,COLA, etc.areconstantovertime[Rosenthal].
Severalavenuesareavailabletoamelioratethesedifficulties.
3.1 Modeltherisk
Thevaluesforranddcanberoundedup(ordown)asappropriatetorepresentafixedrisk
“premium”forthefutureuncertaintyoffinancialandtechnologicaltrends.Abetterapproachis
toaddanadditionalriskcomponentRtothepayasyougoEquation(2).

·ℓ·
··

·
(14)
R Riskcostposedbyuncertaintyoverthefuturevaluesofrandd.
Table9Risknotation
ThegrowinginfluenceofRovertime,reflectiveofincreasinguncertainty,canbecontrolledby
settingtheappropriatevaluefor
ω
R
and
d
R
,theweightingfactorandexpectedannualrateof
changeinriskusedinEquation(14)for
d
,therateofchangefortheaggregatecostof
preservationinEquations(7),(8),(11),and(12).
Insurersprotectthemselvesfromtrulyunexpectedlylargepayoutsbypurchasingreinsurance:in
essence,insuranceagainstexcessiveclaims[Reinsurance].Theadditionalcostintroducedinto
thepricingequations throughtheriskfactorRcanbe
seenasrepresentingareinsurance
“premium”thataccumulatesovertimeinanticipationofmeetingsomefutureunexpectedneed.
Thisisalsosimilartotheideaofa“rainydayfund”thatgovernmentsfundduringperiodsof
budgetarysurplusinanticipationoffutureshortfalls [Rainy].
UCCurationCenter
TotalCostofPreservation(TCP) Page11of20
3.2 Recalibration
Ratherthankeepingkeymodelvariables,
F
,
G
,
r
,and
d
,constantovertime,theycanbe
recalibratedperiodicallyonthebasisofcumulativeexperienceandcurrentknowledge.
RecalibratedpriceswouldapplyforprospectiveProducersonly;retrospectiveProducerswould
remain“lockedin”atthepriceinforceatthetimetheyestablishedtheirrelationshipwiththe
serviceprovider.
3.3 Hybridpricing
Thehybridapproachtopricingdistinguishesbetweenthosecostcomponentsthatareconsidered
easytoquantifyandforecast,andthosethataren’t.Paiduppricingisappliedonlytotheformer;
activitiesassociatedwiththelatterareperformed,ifitall,onapayasyougobasis.
Forexamplepreservation
Interventions,futureactionstakeninresponsetofuture,possibly
unanticipated,conditions,maybeconsideredtofallintothedifficulttoquantifycategory.The
paiduppricewouldthereforeonlyincorporatethecostsforpreservationactivitiesthatcanbe
performedatthetimeofcontentingestviaWorkflowsastheresult
ofContentTypeanalysis,such
asthecreationofnormalizedordesiccated[WAS,Desiccation]derivativeformsofthe
contributedcontent.
Inamoreextremecase,thebaselinepaiduppreservationfunctioncouldberestrictedtobitlevel
preservationonly.Thisservicemoderequiresonlythreecostcomponents:theSystem,Storage
,
andProducer;supportforWorkflowisnotnecessaryasthesinglerequiredworkflowisassumed
tobesubsumedundertheSystem,whileContentTyp es,Monitoring,andInterventionsarenot
necessarysinceonlybitlevelactivitiesareencompassedatthisservicelevel.
3.4 Shortentheterm
Uncertaintyinthepredictivereliabilityofthevalueschosenforranddisproportiontothelength
oftimeunderconsideration;thelongertheterm,thegreatertheuncertainty.Withthisinmind,
itmaybeprudenttoshortenthetermtoonemoreamenabletoreliableforecasting,withthe
possibilityforsubsequent“renewal”ofpreservationservice.Forsomebodiesofcontentafixed
term10year?20years?mayinfactbeanappropriateinitialterm.Shorttomidterm
preservationprovidesanopportunityfo rthecontenttoproveitsvalue,asevidencedbythe
commitmentof
acuratorialadvocatetopayforitssubsequen tpreservation.
3.5 Stochasticmodeling
RosenthalsuggestsusingmoresophisticatedeconomictechniquesbasedonMonteCarlo
simulationtomodelpotentialfuturescenariosstochastically[Rosenthal].
UCCurationCenter
TotalCostofPreservation(TCP) Page12of20
AppendixAPriorwork
Anumberofinternationaleffortshaves tudiedthequestionoflongtermpreservationcosts;most
notably,theNationaalArchiefoftheNetherlandsin2005[Archi ef];theLIFE
2
project(LifeCycle
InformationforELiterature)workonaGenericPreservationModel(GPM)in2008[Ayris,Hole,
LIFE];theKRDSproject(KeepingResearchDataSafe)in2010[Beagrie,KRDS];thePrinceton
DataSpaceinitiativein2010[Goldstein];andtheDanishNationalArchivesandRoyalLibrary
CMDP(CostModelforDigital
Preservation)in2011[Kejser,CMDP].(See[Zell er]forasummaryof
theseandotherrelevantactivities.)
Allofthesemodelsanalyzepreservationcoststhroughoutthefulllifecycleofpreservedassets.
BoththeNationaalArchiefandLIFEprojectworkemployveryfinegrainedanalysisofcost
componentsandarebasedonrepresentative
actualcostsratherthanthenominalcosts
employedbyTCP.TheLIFEmodelincludescostcomponentsforcontentcreationandacqui sition,
whichareconsideredoutofscopeintheTCPanalysis.Furthermore,LIFEscalespreservation
actioncostsbythenumberofexpectedobjects.TCPassumesthatthereisnoper
objectmarginal
cost;instead,marginalcostsareassociatedonlywiththevariousContentTypesofwhichobjects
aremembers.TheKRDSmodelis specifictotheresearchdatalifecycle,butitsfindingsappearto
beapplicabletoothercontexts.UnlikethetheNationaalArchiefandLIFEmodels,KRDSassumes
adiscounting
functionthatannuallydecreasestheaggregatecostofpreservationservice.The
DataSpacemodelisalsobasedonadiscountingfunction,butitsanalysiscoversonlythecosts
associatedwithpreservationstorage,whicharedefinedonapayonce,storeforever(POSF)basis.
TheOAISactivitybasedCMDPworkconcentrateson
postfactomeasuringofpreservationcosts,
ratherthanonforecasting,althoughitispossibleusingtheframework.WhileDataSpaceis
explicitlyconcernedwith“forever”pricing,neitheritnoranyoftheothermodelsassumethe
benefitofanannualinvestmentreturninoffsettingaportionofongoingcosts.Most,ifnot
all,of
theindividualcostcomponentsofthesemodelscanbemappedtoOAISenvironmentaland
functionalentities[OAIS]orTRACcriteria[TRAC],facilitatingcommonpointsofreferenceand
comparison.
Anumberofinstitutions,bothcommercialandnonprofit/academic,nowofferlongterm
preservationservices.(Oratleastpersistentstorage;
theimportantdistinctionbetweenthetwo
isnotalwaysapparentfromthedescriptionoftheseserviceofferings.)Carboniteis
representativeofcommercialpreservationserviceofferings[Carbonite].Itscostis$599/yearfor
upto500GB,withfurtherincrementsof100GBavailablefor$89/year.TheLifeTimeLibraryat
theUniversityof
NorthCarolinaoffersstudentspermanentstorage“andassociatedservices”for
250GB,apparentlywithnoassociatedfees[UNC].TheUSCDigitalRepositoryoffersapaidup
licensefor20yearsofpreservationservicefor$1000/TB[USC].Ingeneral,however,little
informationisavailableexplainingthebasisoftheirbusinessmodels.

UCCurationCenter
TotalCostofPreservation(TCP) Page13of20
AppendixB Powerseries
Thesolutionforageneralpowerseries,
·


,canbederivedasfollows:
· ·
·

(15)
Equation(15)canbescaledbyafactorofx:
·
· ·
  ·

·
(16)
SubtractingEquation(16)from(15)yields:
·
··
·

·

  ·
(17)
·
·
(18)
·
1
· 1  
(19)
Thus,dividingEquation(19)by1–
x
:
·
1
1
where
x
 1
(20)
Theequationfor
p
x
asaninfiniteseriesisderivedbyobservingthat
x
T
inEquation(20)
approaches0as
T
approaches,solongas0≤
x
 1.
lim

0
1
(21)
lim

1
where0 
x
 1
(22)
VariousotherequationsintheTCPmodelsarederivedbysubstitutingtheappropriatetermsfora
and
x
inEquations(20)and(22).

UCCurationCenter
TotalCostofPreservation(TCP) Page14of20
AppendixC Payasyougopricemodel
Thecumulativediscounte dpayasyougopriceover
T
yearsisbasedonthesumofa
geometricallydecreasingseriesofpayments,eachproportionallydiscountedfromGbyafactorof
1–
d
(seeFigure2):
,
·
1
·1
·1

(23)
ThesolutionforEquation(23)issimplythegeneralsolutionforafinitepowerseriesgivenby
Equation(20)with
G
substitutedfor
a
and1–
d
for
x
.
,·
1
1
where
a
G
and
x
 1
d
(24)
Substitutingfor
a
and
x
andsimplifyingproducesthecumulativediscountedpayasyougoprice
over
T
years.
,·
11
where
d
 0
(25)
Likewise,thefinitepowerseriesalsohasasolutionforthecasewheretimeisarbitrarilylargeas
longasitsatisfiestheconvergencecriterion, 0 
x
 1.
∞,
1
wherea=Gandx=1d
(26)
Substitutinginfor
a
and
x
andsimplifyingproducesthecumulativediscountedpayasyougo
pricefor“forever.”

,
where0  1–
d
1
(27)

UCCurationCenter
TotalCostofPreservation(TCP) Page15of20
AppendixD Discountfactor
Thevariable
d
inEquations(7),(8),(11),and(12)isacompositevalueorfunctionthatsumsthe
individualcomponentspecificchangesinquantityandunitcost,weightedbythecomponent’s
proportionatecontributiontotheoverallcost(seeAppendixE).

·
 
·
·

·
·

·
·

·
·
 
·
·

·

(28)
where
·
(29)
·
·
(30)
ℓ·
·
(31)
·
·
(32)
·
·
(33)
·
(34)
·
(35)
d
Annualpercentagerateofchangeinaggregatepreservationservicecost.
A
WeightingfactorfortheSystem component.
d
A
AnnualpercentagerateofchangeinfixedcostofthearchivalSystem.
W
WeightingfactorfortheWorkflow component.
d
m
AnnualpercentagerateofchangeinthenumberofsupportedWorkflows.
d
W
AnnualpercentagerateofchangeinthenominalunitcostofaWorkflow.
C
WeightingfactorfortheContentType component.
d
AnnualpercentagerateofchangeinthenumberofsupportedContentTypes.
d
C
AnnualpercentagerateofchangeinthenominalunitcostofaContentType.
M
WeightingfactorfortheMonitoring component.
d
j
AnnualpercentagerateofchangeinthenumberofsupportedMonitoring
activities.
d
M
AnnualpercentagerateofchangeinthenominalunitcostofaMonitoring
activity.
V
WeightingfactorfortheInte rvention component.
UCCurationCenter
TotalCostofPreservation(TCP) Page16of20
d
i
AnnualpercentagerateofchangeinthenumberofsupportedInterventions.
d
V
AnnualpercentagerateofchangeinthenominalunitcostofanIntervention.
S
WeightingfactorfortheStorage component.
d
k
AnnualpercentagerateofchangeinthenumberofStorage unitsconsumed.
d
S
AnnualpercentagerateofchangeinthenominalcostofaunitofStorage.
O
WeightingfactorfortheOversight component.
d
O
AnnualpercentagerateofchangeinthenumbercostofaunitofOversight.
Table11Changeinpreservationservicecostnotation
SincetheSystemandOversightarefixedcosts,thenumeratorsofEquations(29)and(35)only
havetheunitcostterm;allotherweightingfactorshavebothnumberofunitsandunitcost
terms.SincethecostsforSystem,Workflows,
ContentTypes,Monitoring,Interventions, and
Oversightarecommongoods,theyareapportionedacrossallnProducers,asreflectedinthe
denominatorsofEquations(29)(33)and(35);StorageisalwaysspecifictoagivenProducer,so
thereisnonterminEquation(34).
Notethatallof
thediscountfactorsarerealratesofchange,incorporatingtheeffects,ifany,of
inflation.

UCCurationCenter
TotalCostofPreservation(TCP) Page17of20
AppendixE Paiduppricemodel
Likethepayasyougoprice,thepaiduppriceover
T
yearsisbasedonthesumofageometrically
decreasingseriesofpayments(seeFigure3).Thepresentvalue
F
ofthefutureseriesof
paymentsusingthebasecostGdecliningannuallyby1–d,andanominalinvestmentreturn
r

compoundedmonthlyis:
,,
·
·
1
1
 ·
·
1
1
·
·
1
1
…
·
·
1

1
(36)
Theeffectiveinvestmentreturne,incorporatingtheeffectofmonthlycompounding,is[Tucker]:
1
1
12

(37)
Eachdecliningterm,·
1
,representingafuturepaymentisconvertedtoitspresentvalue
byadjustingforthemonthlycompoundedtimevalueofmoney,thatis,scalingby
·
1

,
andtheresultsaresummedtoproducethetotalpresentvalue.Rearrangingthetermsof
Equation(36)gives:
,,
1
·
·1
1
1
1
1
…
1

1

(38)
ThesolutiontoEquation(38),issimplythegeneralsolutiontothepowerseriesgivenbyEquation
(20)whereais

·
andxis


.
,, ·
1
1
where
1
·
and
1
1
(39)
Substitutingfor
a
and
x
andsimplifyingproducesthepaiduppricefor
T
years.
,, ·
·
1
1
1
·
(40)
Likewise,thefinitepowerseriesalsohasasolutionforthecasewheretimeislargeaslongasit
satisfiestheconvergencecriterion,0 
x
 1.
∞,,
1
where
1
·
and
1
1
(41)
Substitutingfor
a
and
x
andsimplifyingproducesthepaiduppricefor“forever.”
UCCurationCenter
TotalCostofPreservation(TCP) Page18of20
∞,,
·

where0
1
1
1
(42)
UCCurationCenter
TotalCostofPreservation(TCP) Page19of20
References
[Archief] NationaalArchief(2005),CostsofDigitalPreservation,
<http://www.nationaalarchief.nl/sites/default/files/docs/kennisbank/codpv1.pdf
>,accessed
20110902.Seealso,
<http://www.nationaalarchief.nl/sites/default/files/docs/kennisbank/kostenmodel_in_excel_
versie_1.2_final.xls>,accessed20110902.
[Ayris] Ayris,Paul,RichardDavies,RoryMcLeod,RuiMiao,HelenShenton,andPaulWheatley(2008),
LIFE2FinalProjectReport,<http://eprints.ucl.ac.uk/11758
>.
[Beagrie] Beagrie,Neil,BrianLavoie,andMatthewWoollard(2010),KeepingResearchDataSafe2
<http://www.jisc.ac.uk/media/documents/publications/reports/2010/keepingresearchdatasaf
e2.pdf>.
[Carbonite] Carbonite,Inc.(2011),OnlineBackupData&FileBackupSoftware|Carbonite
<http://www.carbonite.com/en/default.aspx
>.
[CMDP] DanishNationalArchives/DanishRoyalLibrary(2011),CostModelforDigitalPreservation
(CMDP)<http://www.costmodelfordigitalpreservation.dk/
>,accessed20110902.Seealso,
<http://www.costmodelfordigitalpreservation.dk/contact/cmdp1‐‐‐preservationplanning
anddigitalmigrations/documentation/Kostenmodel_in_Excel_versie_1.2_%28final%29.xls>
and<http://www.costmodelfordigitalpreservation.dk/contact/cmdp2‐‐‐ingestandarchival
storage/documentation/Tool_Version_2.xls>.
[Damodaran]Damodaran,Aswath(2002),InvestmentValuation:ToolsandTechniquesforDeterminingthe
ValueofAnyAsset(2nded.;NewYork:Wiley).
[Dappert] Dappert,Angela(2011),“IsIst[sic]possibletopredictthecostoflongtermdigital
preservation?”OpenPlanetsFoundationblog,
<http://www.openplanetsfoundation.org/blogs/20110803istpossiblepredictcost
longtermdigitalpreservation>,accessed20110902.
[Desiccation]Phillips,MarkEdward,andDanielGelawAlemneh(2011),“DataDesiccation:Facilitatinglong
termaccess,use,andreuseofETDs,”Proceedingsofthe14thInternationalSymposiumon
ElectronicThesesandDissertations,CapeTown,2011
<http://digital.library.unt.edu/ark:/67531/metadc67625/
>.
[Goldstein] Goldstein,SergeJ.,andMarkRatliff(2010),DataSpace:AFundingandOperationalModelfor
LongTermPreservationandSharingofResearchData,
<http://arks.princeton.edu/ark:/88435/dsp01w6634361k
>,accessed20110902.
[Hole] Hole,Brian,LiLin,PatrickMcCann,andPaulWheatley(2010),“LIFE
3
:Apredictivecostingtool
fordigitalcollections,”7thInternationalConferenceonPreservationofDigitalObjects
(iPRES2010),Vienna,September1924,2010
<http://www.life.ac.uk/3/docs/Ipres2010_life3_submitted.pdf
>,accessed20111208.
[Kejser] Kejser,UllaBøgvad,AndersBoNielsen,andAlexThirifays(2011),“Costmodelfordigital
preservation:Costofdigitalmigration,”InternationalJournalofDigitalCuration6(1):255267
<http://www.ijdc.net/index.php/ijdc/article/viewFile/177/246
>,accessed20111208.
[Kosten] NationaalArchief(2011),Kennisbankdigitalbewaren:Kosten
<http://www.nationaalarchief.nl/kennisbanken/digitaal
UCCurationCenter
TotalCostofPreservation(TCP) Page20of20
bewaren?auteur=All&onderwerp=28>,accessed20110902.
[KRDS] CharlesBeagrie(2011),KeepingResearchDataSafe(KRDS)
<http://www.beagrie.com/krds.php
>,accessed20110902.
[LIFE] LIFEproject(2011),LIFE:LifeCycleInformationforELiterature<http://www.life.ac.uk/
>.
[Neftci] Neftci,SalihN..(2008),PrinciplesofFinancialEngineering(2nded.;Amsterdam:Elsevier)
<http://www.bizknowledge.info/Mybooks/AcademicPressPrinciplesofFinancial
Engineering2nd.pdf>.
[OAIS] ISO14721(2003),SpacedataandinformationtransfersystemsOpenarchivalinformation
systemReferencemodel.Seealso,CCSDS650.0B1(2002),BlueBook
<http://public.ccsds.org/publications/archive/650x0b1.pdf
>.
[Presto] PrestoPRIME(2011),StoragePlanningTool,version0.1.4<http://prestoprime.it
innovation.soton.ac.uk/planningtool/>,accessed20110902.
[Rainy] BrookingsInstitute(2010),StateandLocalTaxPolicy:whatarerainydayfundsandhowdo
theywork?<http://www.taxpolicycenter.org/briefingbook/statelocal/fiscal/rainyday.cfm
>,
accessed20120120.
[
Reinsurance] ReinsuranceAssociationofAmerica(2007),IntroductiontoPropertyandCasualtyReinsurance
<http://www.reinsurance.org/files/public/07FundamentalsandGlossary1.pdf
>,accessed2012
0119.
[Rosenthal] Rosenthal,David(2011),“Modelingtheeconomicsoflongtermstorage,”dshr’sblog,
September27,2011<http://blog.dshr.org/2011/09/modelingeconomicsoflongterm
storage.html>.
[TRAC] OCLC/CenterforResearchLibraries(2007),TrustedRepositoryAudit&Certification:Criteria
andChecklist<www.crl.edu/sites/default/files/attachments/pages/trac_0.pdf
>.
[Tucker] Tucker,WilliamR.(2000),Effectiveinterestrate(EIR),BankakademieMicroBanking
CompetenceCenter.<http://www.docstoc.com/docs/3649810/WilliamRTuckerEffective
InterestRateEIREIRPeriodicInterest>,accessed20120406.
[UNC] UniversityofNorthCarolina(2011),LifeTimeLibrarythoughttobethefirstnationwide,
September15,2011<http://sils.unc.edu/news/2011/SILSlifetimelibrary
>.
[USC] UniversityofSouthernCalifornia(2011),USCDigitalRepository:Pricing
<http://repository.usc.edu/pricing/
>,accessed20111208.
[WAS] Kunze,John(2005),“WebArchivingService(WAS):AnNDIIPPCollaboration:
LC/CDL/UNT/NYU,”DLFFallForum,Charlottesville,November79,2005
<http://old.diglib.org/forums/fall2005/presentations/kunze200511.pdf
>.
[Wheatley] Wheatley,Paul,P.Bright,andRoryMcLeod(2008),LIFEGenericPreservationModelv.1.4
<http://discovery.ucl.ac.uk/14128/1/14128.xls
>.
[Zeller] Zeller,JeanDaniel(2010),“Costofdigitalarchiving:Isthereauniversalmodel?”8thEuropean
ConferenceonDigitalArchiving,Geneva,April2830,2010
<http://regarddejanus.files.wordpress.com/2010/05/costsdigitalarchiving_jdz_eca2010.pdf
>,
accessed20110902.