Distributed systems NPTEL lecture slide notes

Vvvvvgfgggghhhhgvhvvvbbbbbbbb

1. DistributedSystems
Prof.RajivMisra
DepartmentofComputerScienceandEngineering
IndianInstituteofTechnology,Patna
Lecture-01
IntroductiontoDistributedSystems
SothisisthefirstlectureIntroductiontotheDistributedSystems.
(ReferSlideTime:00:21)
Inthisparticulardistributedsystemlectureintroduction,wearegoingtodiscussaboutthe
differentrequirementsofadistributedcomputingsystemsandwearegoingtodiscuss,what
arethedifferenttopicswearegoingtocoverthetextbooksandsoonsoforth.
So,beforethatletusbeginwiththepreface.So,theexplosivegrowthofdistributed
computingsystemsmakesunderstandingimperativeyetdifficultbecauseofuncertainties
introducedbytheasynchrony,limitedlocalknowledge,andpartialfailures.Thenaturesolves
itperfectly,suchasflockofbirdswherethesebirdsarethemobileagentstheycommunicate
witheachothertoachieveacommongoal.
However,inthefieldofdistributedcomputingprovidingalltheseintricacies,thatis
asynchrony,limitedlocalknowledge,partialfailures.Tounderstandthiscoursewillprovidea
theoreticalunderpinningfordesignandanalysisofmanydistributedsystems,suchasandthe
conceptssuchascommunication,coordination,synchronizationanduncertaintytothelower

2. boundstechniques.Thesetogetherwillbediscussedinthepartofthecourse.Andthis
particularcoursewillbequiteusefulasfarasthedifferentapplicationsareconcerned.
Thecoursestructureofadistributedsystemgoeslikethis.
(ReferSlideTime:02:00)
So,thisparticularcourseifyouseeisdividedintotwodifferentparts.Thefirstpartisthe
perspectivefromthesystemsperspectivedistributedsystems.Thesecondpartiscalled
distributedsystemsfromalgorithmsperspective.So,thealgorithmswhichwillrunonthis
modelofadistributedsystemarerequiredtobeunderstoodhere.So,modelalsoisrequired
tobeunderstood.Andthenthealgorithms:howtodesignthesealgorithms,howtoanalyzeit
anddifferentintricaciesofthisalgorithmdesigninthisparticularproblemsetting.
So,themaintopicswhichwearegoingtofocusonfromalgorithmsperspectivemeanshow
tobuildthespanningtreesusingfloodingalgorithms,thentheleaderelectionalgorithm.
Thesearebasicallymostoftheimportantalgorithmicdesigntechniquestheyarethebasic
buildingblocksofthedistributedsystems.Fromsystemsperspectivewearegoingtocover
uptheglobalstaterecording,mutualexclusion,consensus,sharedmemory,checkpointing,
rollback,distributedhashtable.Andthecasestudiesofadistributedsystemswhichwewill
coverhereinthispartofthecoursestructureispeertopeer,Googlefilesystem,HDFSand
introductiontothespark.

3. (ReferSlideTime:03:19)
Withthisparticularcourse,wewillusetwotextbooks.Thefirstwewilldealaboutthe
systemsperspective,thefirstonementionedhereastheauthorsKshemkalyaniandSinghal.
TheothertextbookswilldealwiththealgorithmsperspectiveandthatisbyJenniferWelch.
WehavethereferencebookalsothatisdistributedalgorithmsbyNancyLynch.
(ReferSlideTime:03:47)
Letusbeginwiththedefinitionofadistributedsystem.Distributedsystemisacollectionof
independententitiesthatcooperatetosolveaproblemthatcannotbesolvedindividually.So,
basicallyitisnothingbutacollectionofcomputers.Thus,thisparticularcollectiondonot

4. shareacommonmemoryordonothaveacommonphysicalclock,andtheonlywaytheycan
communicateisthroughthemessagepassingandforthattheyrequireacommunication
network.
Thecomputersusedhereindistributedsystemsaresemi-autonomousandtheyareloosely
coupledwhiletheycooperatetoaddresstheproblemcollectively.
(ReferSlideTime:04:33)
So,beforeweunderstandinmoredetailaboutthedistributedsystems,letushavesome
propertiesofdistributedsystemstokeepinamindatthispointoftime.So,heterogeneityis
oneoftheproperties,becauseherethesystemcomprisesofdifferentcomputersautonomous
computersandtheymaybeheterogeneoushavingheterogeneoushardwareandsoftware
components.
Theconcurrencyisanotherpropertyofadistributedsystem,shareddataisalsoanother
propertyofadistributedsystem.So,noglobalclockisalsooneoftheimportantproperties
andinterdependenciesarethere.Theyhaveinterdependentcomponentsdependoneach
other.

5. (ReferSlideTime:05:17)
Now,tounderstandthedistributedsystemfromthesystemperspective,letusseethis
particularfigureordiagram.Inthisparticulardiagram,youcanseethecomputersare
autonomouscomputerspresentedasprocessormemorywiththeoperatingsystemand
basicallythecommunicationprotocolstack.
And,thisparticularthesedifferentcomputerstheycancommunicatethroughthenetworkthat
isthecommunicationnetwork.Now,asfarasthesoftwareisconcernedwhichwillbuilda
distributedsystemthisparticularsoftwareiscalledbasicallythemiddlewareandthepartof
thesemiddlewareisbasicallythesoftwarewhichrunsoneachcomputertheyarecalled
software’s;theyarewrittenasthesoftwarecomponents.
So,basicallythisparticulardistributedsystemsoftwarewillbasicallyusetheexisting
computerstheiroperatingsystemandunderlyingcomputernetworkandtheyrunthepartof
themiddlewareandtogetherthiswillformadistributedsystem.So,middlewarewillbindthe
distributedsystem.

6. (ReferSlideTime:06:25)
So,againfurtherexplainthedistributedsystemconnectsautonomousprocessorsby
communicationnetwork.Andthesoftwarecomponentthatrunoneachofthecomputersuse
thelocaloperatingsystemandnetworkprotocolstack.Thedistributedsoftwareistermedas
middleware.
Thedistributedexecutionistheexecutionoftheprocessesacrossthedistributedsystemto
collectivelyachieveacommongoal.Theexecutionisalsosometimestermedasthe
computationorerrorinadistributedsystem.
(ReferSlideTime:06:57)

7. Furthermore;thedistributedsoftwarewhichisalsocalledamiddlewareisdesignedina
layeredarchitecturetosimplifythecomplexityofthedistributedsoftware.Andthisparticular
middlewareorthedistributedsoftwarethatdrivesthedistributedsystem,italsoprovidesthe
concurrencyofheterogeneityattheplatformlevel.
So,inthediagramyoucanseethedistributedapplicationmentionedoverhere.So,this
particulardistributedapplicationwillusethedistributedsoftwarewhichisamiddleware;
middlewarerunsontheoperatingsystemofeachcollectionofcomputersandalsoitwilluse
theunderlyingnetworkprotocolstackforthecommunication.Andthereareseveral
standardsalsoevolvedoveratimeforthisparticularmiddlewareapplication;middlewarefor
thedistributedsoftwaredevelopment.ThatisOMG,CORBA,RPC,DCOM,RMI,MPI,and
soon.
(ReferSlideTime:08:02)
So,thatwastheoverviewofthedistributedsystemfromasystemperspective.Now,weare
goingtotouchuponthemotivationofthedistributedsystem.So,inherentlydistributed
computationthatismanyapplicationssuchasmoneytransferinthebanking,orreachinga
consensusamongthepartiesthataregeographicallydistant,thecomputationisinherently
distributed.So,forthatthemodelthatisdistributedsystemisrequiredforthatcomputation
thatistheapplicationswhichareinherentlydistributed.
Then,nextisthemotivationcalledresourcesharingthesharingoftheresourcessuchas
peripherals,andacompletedatasetandsoonandsoforth;isabasicallythemotivation

8. behindthisbuildingofdistributedsystem.Anothermotivationistoaccessthegeographically
remotedataandresources,suchasbankdatabase,supercomputerandsoon.Reliability:
enhancedreliabilitypossibilityofreplicatingtheresourcesandexecutiontoenhancethe
reliability.Geographicallydistributedresourcesarenotlikelytocrashatthesametime.That
isthemotivationofbuildingthedistributedsystemforthat.
(ReferSlideTime:09:19)
So,reliabilityentailsseveralaspectsinthatcase.So,thesearebasicallytheavailabilitythe
resourcesshouldbeaccessibleatallthetimes.Integritythevalueorobliquethestateofthe
resourceshouldbecorrect,inthefaceofconcurrentaccess,andfault-tolerancetheabilityto
recoverfromsystemfailures.Increasedperformanceobliquecostratiobyaccessing
geographicallyremotedataandresourcesharing.So,thesearebasicallythereliabilitywill
entailtheseaspects.

9. (ReferSlideTime:10:02)
So,basicallyotheradvantageofdistributedsystemisscalability,addingmoreprocessorto
thecommunicationnetworkdoesnotposeabottlenecktothecommunicationnetwork.Then,
thenextadvantageismodularityandincrementalexpandability.So,heretheprocessor
heterogeneousprocessorcanbeaddedwithoutanybottleneckproblems.
(ReferSlideTime:10:18)
Now,wearegoingtodiscussthedesignissuesandchallengesinthedistributedsystem
design.

10. So,fromsystemperspectiveofadistributedsystemdesign,wearegoingtosee;whatarethe
intricaciesandwehavetounderstandthetheoreticalbasisforthisdesign.Anotherthingis
algorithmicperspectiveofitindistributedsystemdesign.Thenextdesignissuesand
challengesarebasedontherecenttechnologyadvancesandalsodrivenbythenew
applicationsandwhichwillbasicallybethemotivationorbethedesignissues.Andalso
becomingchallengesinevolvingthedistributedsystems.
Sofirstly,wearegoingtolookuponthedesignchallengesfromsystemperspectiveof
distributedsystems.
(ReferSlideTime:11:14)
Here,thecomponentswhichareinvolvedhereinthesystemsperspectivearethe
communicationsthatisthecommunicationnetworkwheretheprocessorscanbasically
communicatewitheachotherthroughwhichprocessorscancommunicate.Processessomeof
theissuesinvolvedare:themanagementoftheprocessesandthethreadsattheclientserver;
codemigration;designofsoftwaremobileagents.Synchronizationisthemostimportantpart.
Synchronizationorthecoordinationamongtheprocessesisessential.Mutualexclusionisan
exampleofsynchronization,butmanyotherformsofsynchronization,suchasleader
election,physicalclocks,logicalclocks,globalstaterecordingalgorithms,allrequire
differentformofsynchronizationthatwearegoingtocoverupinthispartofthecoursein
moredetails.

11. (ReferSlideTime:11:59)
Now,anothersystemlevelchallengeisthefaulttolerance.
So,thisfaulttolerancerequiresmaintainingcorrectnessinspiteofthefailuresofalinks,
nodesandprocesses.So,thisparticularfaulttoleranceisbasicallyachievedusingtheprocess
resilience,reliablecommunication,distributedcommit,checkpointingandrecovery,
agreementandconsensus,failuredetection,self-stabilizationthesearesomeofthetechniques
whichwearegoingtocoverup,whenwediscussthedesignfromsystemsperspective.
(ReferSlideTime:12:38)

12. Anothersystemperspectivedesignangleoraspectistransparency.So,transparencyistohide
theimplementationpoliciesfromtheuserandthiscanbeadifferentkindoftransparencies:
thefirstoneiscalledaccesstransparency.Whenithidesthedifferencesaysinthedata
representationondifferentsystemsandlocationtransparencywhenitmakesthetransparency
ofthelocationoftheresources.
(ReferSlideTime:13:13)
Andmigrationtransparencyallowstherelocatingresourceswithoutchangingthename.
Relocationtransparencytheabilitytorelocatetheresourcesastheyarebeingaccessedis
relocationtransparency.Replicationtransmitdoesnotlettheuserbecomeawareofany
replication.
Concurrencytransmitsdealwithmaskingtheconcurrentuseofsharedresourcesfortheuser.
Failuretransparencyreferstothesystembeingreliableandfault-tolerant.Itisnotknownto
theuseratthisatanypointoftime.

13. (ReferSlideTime:13:38)
Now,thatwasthedistributedsystemfromsystemperspective.Now,wearegoingtotouch
uponanotherimportantcomponentofthisparticulardistributedsystemdistributedcomputing
systemthatiscalleddistributedalgorithms.
So,thealgorithmsaretobeevolved.So,wearegoingtocoverupthefundamental
algorithmswhichwillbethebasicbuildingblocksofdevelopingthedistributedapplications.
So,indistributedapplications,distributedsystems,differentcomplexitymeasuresareof
interestsuchas:thetimeandspace.Theywereusedintheclassicalorthesequential
algorithmsaswell,butnowcommunicationisalsoevolved.So,communicationcostisoneof
thecomplexitymeasures.
So,communicationcostincludesthenumberofmessages,sizeofthemessageandnumberof
sharevariables.Andalsoanothercomponentwhichwillbeusedinthecomplexityiscalled
basicallythenumberoffaultyversusnon-faultycomponents.Nowbecauseofthe
complicationsfacedbydistributedsystemtheyleadtotheincreasethescopeofnegative
results,lowerboundsandimpossibilityresults.
So,allthesethingswillbecoveredupinthedistributedalgorithmdesignandthuswewill
discussmoretheseparticulardistributedalgorithmsinthedetails.So,thefundamentalissues
inthedesignofdistributedalgorithmsarethefollowingthreefactors.

14. (ReferSlideTime:15:22)
Theasynchrony:soasynchrony,limitedknowledgeandfailures.Therearethreedifferent
importantfundamentaldesignissuesinthedistributealgorithm.Asynchronyisbasically
absoluteandrelativetimingoftheeventscannotbeknownprecisely.
So,inthisparticularsettinghowthealgorithmsaretobeevolvedordeveloped.Localview
thatisthecomputingentitiescanonlybeawareoftheinformationitacquires,soithasonly
thelocalviewofaglobalsituation.Thirdoneisthefailures.So,thecomputingentitiescan
failindependently,leavingsomecomponentsoperationalwhileothersarenot.So,thesethree
differentfactorstheyaddthemorecomplicationsindesignofthedistributealgorithmand
becomesachallengingtoevolvethedistributealgorithmintheseparticularproblemsetting,
thatisasynchronythatiswearenotknowingtheeventswhenittheyaregoingtooccur.
Localview,wedonotknowthecompletepictureoftheglobalsituationyetwehavetocome
upwithanalgorithm.Failuresmeansthecomponentswhicharebasicallyinvolvedinthat
distributedsystemtheycanfailindependently,andbasicallythisexpectedthatthe
applicationsshouldbasicallykeeponrunninginspiteoffailures.

15. (ReferSlideTime:16:55)
So,distributedcomputingsystemsarestudiedsince1967,startingwithDijkstraandLamport.
Dijkstrain1972gotTuringawardfortheworksonthedistributedalgorithmsanddistributed
systems.LeslieLamportveryrecentlyhasgottheTuringawardforhisworkonbasicallythe
distributedalgorithmsandsystems.
(ReferSlideTime:17:13)
SpecialmentiontotheLeslieLamport,becausemostoftheworkswhateverhehasdonewe
aregoingtocoverupasfarasdistributedsystemsfundamentalsareconcerned.

16. So,LeslieLamportdevisedimportantalgorithmsdevelopformalmodelverificationprotocols
toimprovethequalityofrealdistributedsystems.Fundamentalcontributiontothetheoryand
practice,notablytheinventionsoftheconceptssuchascausalityandlogicalclocks,safety
andliveness,replicatedstatemachines,sequentialconsistencyorsomeofthem.So,Lamport
wasthewinnerof2013thatisTuringawardfordistributedcomputing.
(ReferSlideTime:17:56)
Now,algorithmicchallengesinthedistributedsystem,wearegoingtotouchuponthatis
previouslywehaveseenthedesignchallengesfromsystemperspective.Nowwehavetosee
thealgorithmicchallengesinthedevelopingthedistributedsystemdesigningthedistributed
systems.So,timeandglobalstateinadistributedsystem.So,firstofallthisistheimportant
challengeletusseewhatthisis.Theprocessesinthesystemarespreadacross
three-dimensionalphysicalspace.Anotherdimension,isthetime,hastobesuperimposed
uniformlyacrossaspace.
Thechallengespertaintoprovidingaccuratephysicaltime,becausethereisnocommon
clockandtoprovideavariantofatime,thatiscalledalogicaltime.So,logicaltimeisthe
relativetimeandeliminatestheoverheadofprovidingthephysicaltimeforthedifferent
applications.Andbasicallythelogicaltimebasicallycancapturethelogicandthe
inter-processdependencieswithinthedistributedprogram,andalsotracktherelative
progressateachprocess.

17. So,insteadofphysicalhavingacommonphysicalclockarebasicallyimplementationofa
commonphysicalclockherewearegoingtoseehowthelogicalclockandsolvewithout
havingthephysicalclocktheseparticularproblems.
(ReferSlideTime:19:19)
Theotherproblemotheralgorithmicchallengeisthesynchronizationcoordination
mechanisms.So,theprocessesmustbeallowedtoexecuteconcurrently,exceptwhenthey
needtosynchronizetoexchangetheinformation,thatis,communicateabouttheshareddata;
sosynchronizationessentialforthedistributedprocessestoovercomethelimitedobservation
ofthesystemstate.Thefollowingmechanismsareusedforthesynchronizationandthe
coordination.Firstofallleaderelection:dealswiththeasymmetryofaprocess.Andthen
mutualexclusion:accesstothecriticalresourceshastobecoordinatedthroughthatisdone
throughmutualexclusion.
Thenterminationdetection,thatis,cooperationamongtheprocessestheywillbasicallyable
todetecttherequiredstatetherequiredglobalstatethatiscalledterminationstateandthatis
calledterminationdetectioninadistributedsystem.Thenextimportant;thatmeans,thingis
calledgarbagecollectiondetectingthegarbagerequiresanothercoordination.So,theseare
basicallythesynchronizationandcoordinationmechanismswhichwillbasicallybetheused
upindesigningthedistributedapplications.

18. (ReferSlideTime:20:46)
Anotherthingisanotherimportantnotionisthereliableandthefault-tolerantdistributed
system.So,reliableandfault-tolerantenvironmenthasmultiplerequirementaspect,andthese
canbeaddressbythevariousstrategieswhichwearegoingtocoverupinthispartofthe
course.
Thefirstoneiscalledconsensusalgorithm,secondisthereplicationandthereplica
management,votingandquorumsystems,distributeddatabasesanddistributedcommit,
self-stabilizationsystemandcheck-pointingandrecoveryalgorithm,andfailuredetectors.
So,thesetogether;thesestrategieswillbeabletoprovidethefault-tolerance,thatis,ifthe
componentslikenodes,linksarefailingyethowthedistributedapplicationcanbasically
workonwithoutanydisruptionsandthatisthemainrequirementofthedistributed
motivationofadistributedsystemtohaveareliableinadistributedsystem.
So,basicallyiftheprocessorsorthealgorithmswhicharerequiredareconsensusalgorithm
becauseinthefailures.So,theremainingnon-faultyprocessestheyhavetobasicallycomeup
withtheirconsensusonthevaluesandtheapplicationsarecontinuingtorun.Anotherthingis
calledreplicaandreplicamanagement,becausethereplicasofthatdataisavailable,thenthe
applicationcanrunwithouttheproblemoffailures.
Thenvotingandaquorumisalsoaimportantcriteriaforexample,whensomeofthesystems
orwhensomeoftheimportantsystemisfailed,thenbasicallyamongtheremainingthey
havetoevolvethroughthevotingandquorummechanismtobasicallyrunthoseapplications

19. anddistributeddatabasesanddistributedcommitisalsobasicallyoneoftheimportant
applicationswheretheyhavetodecideamongdiscussionwitheachothertheyhaveamong
thesynchronizationtoseewhetherthecommitwhichistakingplacehastobedoneorhasto
beabortedbytakingthedecisions.
Then,self-stabilizationsystemmeansifthecomponentsarefailingthenhowthesystem
evolvesandhowmuchtimeittakestostabilizeitself.So,thatisself-stabilizationsystems
thatwearegoingtoalsocoverupbrieflyinthispartofthecourse.Check-pointingand
recoverysystemsareveryimportantasfarasthefault-tolerantisconcernedfault-tolerant
meansifthereisafailurehowbasicallytheoperationswhicharedonehastobebasically
withtheminimallosshastobasicallyresumetheiroperations.So,check-pointingand
rollbackrecoveryalgorithms,wearegoingtocoverupinthispartofthecourse.Thenfailure
detectorsarealsoveryimportantifhowtodetectthatthenodesorthelinksarebasicallynot
workingorfailed.
(ReferSlideTime:23:48)
Anotherimportantalgorithmicchallengeisbasicallyformingthegroupcommunication,
multicast,andorderedmessagedelivery.So,therearesomeapplicationswherethegroup
communicationisrequiredso,basicallythisparadigmisalsousefulforthemtodevelopthe
application.

20. (ReferSlideTime:24:03)
Distributedsharedmemoryabstraction;nowthemiddlewareisthesoftwareisthedistributed
systemsoftwarewillbasicallyusethisparticularthisoneabstractioncalleddistributedshared
memory.Although,itisnothavingacommonmemory,butdistributedsharedmemoryis
realizableusingthemessagepassingsystemsthatwearegoingtosee.
(ReferSlideTime:24:26)
Now,theapplicationsofdistributedcomputingsystemsandbasicallythenewerchallenges
are:Mobilesystems:Mobilesystemstypicallyusethewirelesscommunicationwhichis
basedonelectromagneticwavesandutilizesasharedbroadcastmedium.So,mobilesystem

21. isoneoftheapplicationsofadistributedcomputingapplicationwheredifferentelementsare
involvedandtocomeupwiththisparticularserviceandthatiscalledamobileserviceand
thatisanapplicationofadistributedcomputing.Sensornetworkisanotherapplication.
Sensorisaprocessorequippedwithanelectro-mechanicalinterfaceandthatiscapableof
sensingphysicalparameters,suchastemperature,velocity,pressure,humidity,andchemical.
So,thisparticularkindofnodescalledasensornode,ifbasicallydeployedtobasically
monitorthecyberphysicalwillmakeacyberphysicalsystemtomonitoranyphysical
activityoreventforexample,tomonitorthewhetheritishavingavolcaniceruptionorany
othersituation.
So,itishavingalotofuseincyberphysicalsystemanditisakindoflargescaledistributed
systemthatisbasicallyusingtheprinciplesofadistributedsystems.Anotherapplicationhere
iscalledubiquitousorapervasivecomputing:Ubiquitoussystemsrepresenttheclassof
computingwheretheprocessorsembeddedinseamlesslyprovidingthroughtheenvironment
performtheapplicationsoffunctionsinthebackground.Examplesarethesmart
environment;smartenvironmentorsmartbuilding,smartcitiesallareexamplesofa
ubiquitousandpervasivecomputing.
(ReferSlideTime:26:18)
AnotherapplicationisbasicallycalledPeer-to-Peercomputing:Peer-to-Peerrepresentsthe
computingovertheapplicationnetworkwhereallinteractionsamongtheprocessorsareat
thepeerlevels,withoutanyhierarchyamongtheprocessors.Thus,allprocessorsareequal

22. andplayasymmetricalroleinthecomputation.So,peer-to-peernetworksystemsareusedin
providingtheresourcesandservices.Thepeer-to-peernetworkingisnowbasicallyavery
challengingasfarasdevelopingdistributedapplicationsareconcerned.Technicallyalsoitis
quitedifficultand,butthesimplermodellikeclientserverparadigmwhichisnotpurelya
peer-to-peeritisnotpurelydistributed,thisisusedbythedifferentapplicationsbecause
industryfeelscomfortablewithclientservermodel.
Peer-to-peercomputingmodelsfordifferentapplicationsareevolvinganoveraperiodof
timetherecentapplicationwhichisaddediscalledbitcoinandbitcoinisbasedonpeerto
peerdistributedcomputingdesign.Wearegoingtotouchuponlateroninthispartofthe
course.
(ReferSlideTime:27:39)
Distributeddataminingisanotherapplicationofadistributedcomputing.So,distributeddata
miningalgorithmsexaminelargeamountofdatatodetectthepatternsandtrendsinthedata,
tomineorexactusefulinformation.Thetraditionalexampleis:examiningthepurchasing
patternsofthecustomerinordertoprofilethecustomersandenhancetheefficiencyofthe
directedmarketingschemes.
Anotherapplicationofadistributedcomputingisfoundinthegridcomputing:Analogousto
theelectricalpowerdistributiongrid,itisenvisagedthatinformationandcomputinggridwill
becomearealitysomeday.Verysimplystated,idleCPUtimesofthemachinesconnectedto
thenetworkwillbeavailabletotheothers.

23. (ReferSlideTime:28:33)
Anotherapplicationisthesecurityinthedistributed.Thetraditionalchallengesofthe
securityinadistributedsettinginclude:confidentiality;thatmeans,onlyauthorizepersoncan
access,authenticationmeansensurethatthesourcereceivedtheinformationandtheidentity
ofthesendingprocessarebasicallythegenuineandavailabilityisthemaintainingallowed
accessofservicedespitemaliciousactions.
So,securityinadistributedsystemisusedinthepaymentsystems,theonlinepurchasingand
alsotherecentbitcoin,thatis,thedigitalmoneyhowthedigitalmoneyisrealizedlotof
securityalgorithmsareinvolvedinadistributedsetting.So,wearegoingtocoverupthis
aspectalsoandbasicallythisisimportant,becauseinthefinancialmarketthisismostofthis
isoneofthemostimportantfactorsindesigningsuchapplicationsinthedistributedsystems.
So,intheNutshell,whatwehaveseenhereisthedistributedsystemsashavingawide
varietyofapplicationrealworldscenariosandsomeofthemwehavecoveredupinthispart
oftheintroduction.Andthiswillbethebasisandofunderstandingorpinpointingor
underpinningtheintricaciesthetheoreticalandintricaciestounderstandthereasoninghow
thethingsarebeingdesignedandevolvedandalsowecanverifyitscorrectness,thatitis
correctlyworkingandalsotounderstanditscontributionitisrequiredtobefamiliarwiththe
fundamentalprinciples.

24. (ReferSlideTime:30:17)
So,Itoldyouaboutthis.So,thislecturefirstcategorizesthedistributedsystemsandthe
distributedalgorithmsbylookingatvariousinformaldefinitions.Thedesignissuesandthe
challengesbasedontheoreticalandthesystemsaspects.
Intheupcominglecture,wewilltrytogiveaninsightonthedetailedconceptthatwillgivea
goodunderstandingofthefurtherdetails.
Thankyou.