NoSQL in EnterpriseWord文档格式.docx
- 文档编号:5721427
- 上传时间:2023-05-05
- 格式:DOCX
- 页数:16
- 大小:335.19KB
NoSQL in EnterpriseWord文档格式.docx
《NoSQL in EnterpriseWord文档格式.docx》由会员分享,可在线阅读,更多相关《NoSQL in EnterpriseWord文档格式.docx(16页珍藏版)》请在冰点文库上搜索。
spaperon
Dynamo.LastoneyearorsowesawahugeNoSQLmomentumthroughexplosionofmorethan
25products/solution
inthisspacealongwiththeincreasingmindshareacrossdifferentcornersoftheindustry.InthatpretextrecentlyIwasthinkingtotakeadeepdiveonthistoevaluatehowexactlymyclientscangetbenefitedoutofthisNoSQLmovement.Morethanthat,Iwantedtofindoutwhetherthisistherighttimeforenterprisestogiveaseriousthoughtaboutstartingadoptionofthesame.
AquickrecaponwhatisNoSQL
Likemanyotherswhofollowthisspace,IdonotlikethesenseofopposingSQLinherentlyassociatedwiththetermNoSQL.NeitherIlikethecurrentimprovisationofthename,'
NotOnlySQL'
.TomewhatallwearetalkinghereisnotaboutwhethertouseSQLornot.(Onthecontrary,onemaystilldecidetouseSQLlikequeryinterface(withoutsupportforjoin,etc.)tointeractwiththesedatabasesjusttomanagethedevelopmentscalabilityandmaintainabilitywithexistingresourceskills.).
ThismovementisratheraboutfiguringoutwhataretheotherefficientoptionsofstoringandretrievingdatainsteadofblindlytakingtheRDBMSapproachasdefactoforanythingandeverything.Andhencetome'
NonRelationalDatabases'
isabetternametosummarizetheidea.
Whatevermaybethename,thescopeof'
islittleopen(andnegationoriented)witha'
catchall'
typeconnotationimplicittoit.Thatinturnmakespeople(especiallytheenterprisedecisionmakers)confusedaboutwhatisthereandwhatnotandmoreimportantlywhyitmakessenseforthem.
Keepingthatinmind,hereItrytocapturethespiritof'
throughthebelowmentionedcharacteristics.
The'
aretheoneswhich
1.Logicallymodeldatausinglooselytypedextensibledataschema(Map,ColumnFamily,Document,Graphetc)insteadofmodelingdataintuplesfollowingfixedrelationalschema.
2.Designedforhorizontalscalingthroughdatadistributionmodelacrossmultiplenodesabidingbyprinciplesof
CAPtheorem(ensuringthatanytwoofConsistency,AvailabilityorPartitionabilityareachieved).Thiscomesalongwithnecessarysupportformultipledatacentersanddynamicprovisioning(transparentlyadding/removinganodefromaproductioncluster),alaElasticity.
3.Canpersistdataeitherindiskormemoryorboth;
sometimesinpluggablecustomstores.
4.Supportvarious'
Non-SQL'
interfaces(typicallymorethanone)fordataaccess.
Thevariationsaroundthesefourcharacteristics(LogicalDataModel,DataDistributionModel,DataPersistenceandInterfaces)of'
areverywellcoveredinsomeoftherecent
articles
widelyavailableovertheInternet.SoInsteadofdetailingthesameIsummarizethekeyaspectswithsomeexamplesforaquickreference–
Interfaces–REST(HBase,CouchDB,Riak,etc.),MapReduce(HBase,CouchDB,MongoDB,Hypertable,etc.),Get/Put(Voldemort,Scalaris,etc.),Thrift(HBase,Hypertable,Cassandra,etc.),LanguageSpecificAPIs(MongoDB).
LogicalDataModels–Key-Valueoriented(Voldemort,Dynomiteetc.),ColumnFamiliyoriented(BigTable,HBase,Hypertableetc.),Documentoriented(CouchDB,MongoDBetc.),Graphoriented(Neo4j,Infogridetc.)
DataDistributionModel–ConsistencyandAvailability(HBase,Hypertable,MongoDBetc),AvailabilityandPartitionality(Cassandraetc.).
ConsistencyandPartitionabilityisacombinationwhereAvailabilityofsomeofthenon-quorumnodesiscompromised.Interestinglynoneofthe'
NonRelationalDatabase'
todaysupportsthiscombination.
DataPersistence–MemoryBased(e.g.Redis,Scalaris,Terrastore),DiskBased(e.g.,MongoDB,Riaketc.),CombinationofbothMemoryandDisk(e.g.,HBase,Hypertable,Cassandra).Thetypeofstoragegivesagoodideaofwhattypeofusecasesthesolutioncancaterto.However,inmostofthecasespeoplefindthatthecombinationbasedsolutionisthebestone.Theycatertothehighperformancethoughinmemorydatastoreandalsoensuredurabilitybystoringthedataintodiskafterenoughwriteshavehappened.
HowdoesitfitinEnterpriseIT
Intoday'
senterprisesnotallusecaseslendthemselvesintuitivelytoRDBMS,neithertheyneedthestrictnessofACIDproperty(especiallytheConsistencyandIsolation).Gonearethedaysof80sand90swheremostofthedatastoredinanorganizationdatabaseswerestructured,hadtobegeneratedandaccessesincontrolledmannerandwere'
records'
ofbusinesstransactions.Unarguablythosetypesofdataarestillthereandwillcontinuetobethereandshouldalwaysbemodeled,storedandaccessedusingRDBMS.Butwhathappenstothelargevolumeofuncontrolled,unstructured,informationorienteddataexplosionhappenedinenterprisesinlast15yearswiththeadventofweb,digitalcommerce,socialcomputingetc?
Enterprisesreallydon'
tneedRDBMStostoreandretrievethem,asthecorecharacteristicsofRDBMSdonotfitwiththenatureandusageofthisdata.
TheabovefiguresummarizesemergingpatternsinInformationManagementintoday'
swebcentricenterprises.Andthe'
arebetterchoiceforhandlingthesetrends(comparedtoRDBMSsolutions)giventheirsupportforunstructureddata,horizontalscalabilitythroughpartitioning,highavailabilitysupportetc.
Herearesomeexamplesofusecasessupportingthepoint–
LogMining
–ServerLogs,ApplicationsLogs,UserActivityLogsgetgeneratedinmultiplenodesofacluster.ForproductionproblemsolvingLogminingtoolsarehandywhichcanaccesslogsacrossservers,relatethemandanalyzethem.Customsolutioncanbebuilteasilyforthisusing'
SocialComputingInsight–Manyenterprisestodayhaveprovidedtheirusers(Internalusers,Customers,Partners)abilitytodosocialcomputingthroughmessageforums,blogsetc.Miningthoseunstructureddatatheyarefindingofutmostimportancetogetanideaofusermindsharetofurtherimprovetheservices.Useof'
isaperfectlygoodfitforaddressingthisneed
ExternalDataFeedIntegration–Manycasesenterprisesneedtoconsumewiththedatacomingfromtheirpartners.Obviously,evenafternumberofdiscussionsandnegotiations,enterpriseshavelittlecontrolontheformatofthedatacomingtothem.Also,therearemanysituationswherethoseformatschangeveryfrequentlybasedonthechangesinbusinessofthepartners.'
canbeusedveysuccessfullytosolvethisissuewhiledeveloping/customizingaETLsolution.
HighVolumeEAI
–MostoftheenterpriseshaveheavyvolumetrafficflowingthroughtheirEAIsystem(eitherproductbasedorcustomdeveloped).ThesemessagesflowingthroughtheEAIneedtobetypicallypersistedforreliabilityandauditpurpose.Again'
canbegoodfitasunderlyingdatastoreforthisscenariogiventhevariationindatastructureofthesourceandtargetsystemsaswellasgivenhevolumeinquestion.
Frontendorderprocessingsystems–Giventheexplosionofdigitalcommercethevolumeoforders,applications,servicerequestsflowingthroughdifferentchannelstothesystemsofRetailers,BankersandInsuranceproviders,EntertainmentServiceproviders,Logisticprovidersetc.isenormous.Alsoowingtotherestrictionsandbehaviorpatternsassociatedwithdifferentchannels,thestructuresusingwhichtheinformationiscapturedtypicallylittledifferentineachcasesandneedsdifferenttypeofrulesimposed.Ontopofthat,mostoftheserequestsdatadon'
tneedimmediateprocessingandreconciliationatthebackend.Ratherwhatneededisthattheserequestsneedstobecapturedwithoutanyinterruptionwheneverenduserwantstoputthisforwardfromanywhereacrosstheworld.Laterontypicallyareconciliationsystemupdatesthemtothesourceoftruthbackendsystemsandupdatetheenduserontheorderstatus.Thisscenarioisanotherone,where'
canbeusedforinitiallystoringtheinputsfromendusers.Thisscenarioperfectlylendstowardsuseof'
giventhecharacteristicsofhighvolume,differencesininputdatastructureandacceptabilityof'
EventualConsistency'
duringthereconciliation.
EnterpriseContentManagementService
–ContentManagementisnowusedenterprisewideacrossdifferentfunctionalgroupsSales,Marketing,Retail,HRforthevariouspurposes.Andmostofthetimethechallengesarefacedbyenterprisestobringtogetherrequirementofdifferentgroupsinacommoncontentmanagementserviceplatformintermsofdifferenceinmetadatastructure.'
isagoodfittosolvethisproblemalso.
MergerandAcquisition–EnterprisesfacehugechallengesduringM&
Aastheyneedtoconsolidatesystemscateringtosamefunctions.'
canbeusedtosolvethisproblemeithertoquicklyputtogetheratemporarycommondatastoreorevenarchitectingthefuturedatastore,whichcanaccommodatestructureofexistingcommonapplicationsofmergingcompanies.
Buthowexactlywecanarticulatethebusinessbenefitsofusingthe'
overtraditionalRDBMSsolutio
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- NoSQL in Enterprise
![提示](https://static.bingdoc.com/images/bang_tan.gif)