Ukwakha indawo yokugcina impahla kuwumsebenzi omude futhi onzima.
Okuningi empilweni yephrojekthi kuncike ekutheni imodeli yento kanye nesakhiwo sesisekelo kucatshangelwa kahle kanjani ekuqaleni.
Indlela eyamukelwa ngokuvamile iye yaba futhi isalokhu iyizinhlobonhlobo ezihlukahlukene zokuhlanganisa isikimu senkanyezi nefomu lesithathu elivamile. Njengomthetho, ngokwesimiso: idatha yokuqala - 3NF, imibukiso - inkanyezi. Le ndlela, ehlolwe isikhathi futhi esekelwe inani elikhulu locwaningo, iyokuqala (futhi ngezinye izikhathi ukuphela) into efika engqondweni yochwepheshe we-DWH onolwazi lapho ecabanga ukuthi inqolobane yokuhlaziya kufanele ibukeke kanjani.
Ngakolunye uhlangothi, ibhizinisi ngokujwayelekile kanye nezidingo zamakhasimende ikakhulukazi zivame ukushintsha ngokushesha, futhi idatha ivame ukukhula kokubili "ngokujulile" kanye "nobubanzi". Futhi kulapho kuvela khona ukungalungi okukhulu kwenkanyezi - okulinganiselwe ukuguquguquka.
Futhi uma empilweni yakho ethule futhi ethokomele njengonjiniyela we-DWH kungazelelwe:
- umsebenzi waphakama "ukwenza okungenani okuthile ngokushesha, bese sizobona";
- kwavela iphrojekthi ethuthuka ngokushesha, ngokuxhumeka kwemithombo emisha nokusebenza kabusha kwemodeli yebhizinisi okungenani kanye ngesonto;
- sekuvele ikhasimende elingazi ukuthi uhlelo kufanele lubukeke kanjani nokuthi yimiphi imisebenzi okufanele yenze ekugcineni, kodwa selilungele ukuzama futhi licolise ngokungaguquki umphumela eliwufunayo kuyilapho lihlala lisondela kulo;
- Umphathi wohlelo waphendula ngezindaba ezinhle: “Manje siyashesha!”
Noma uma unentshisekelo yokuthola ukuthi ungazakha kanjani ezinye izindawo zokugcina - wamukelekile ekusikeni!
Kusho ukuthini "ukuguquguquka"?
Okokuqala, ake sichaze ukuthi yiziphi izakhiwo isistimu okufanele ibe nazo ukuze ibizwe ngokuthi “ivumelana nezimo”.
Ngokwehlukana, kufanelekile ukusho ukuthi izakhiwo ezichazwe kufanele zihlobane ngokuqondile uhlelo, hhayi inqubo ukuthuthukiswa kwayo. Ngakho-ke, uma ubufuna ukufunda nge-Agile njengendlela yokuthuthukisa, kungcono ukufunda ezinye izindatshana. Isibonelo, khona lapho, ku-Habré, kunezinto eziningi ezithokozisayo (njenge и , futhi ).
Lokhu akusho ukuthi inqubo yokuthuthukisa kanye nesakhiwo senqolobane yedatha akuhlobene ngokuphelele. Sekukonke, kufanele kube lula kakhulu ukuthuthukisa inqolobane ye-Agile yokwakhiwa kwe-agile. Kodwa-ke, ekusebenzeni, kaningi kunezinketho ngokuthuthukiswa kwe-Agile kwe-DWH yakudala ngokusho kukaKimbal kanye neDathaVault - ngokusho kwe-Waterfall, kunokuhlangana okujabulisayo kokuguquguquka ngezindlela zayo ezimbili kuphrojekthi eyodwa.
Ngakho-ke, yimaphi amakhono okufanele isitoreji esivumelana nezimo sibe nawo? Kunamaphuzu amathathu lapha:
- Ukulethwa kusenesikhathi kanye nokushintsha ngokushesha - lokhu kusho ukuthi umphumela webhizinisi lokuqala (ngokwesibonelo, imibiko yokuqala yokusebenza) kufanele utholakale kusenesikhathi, okungukuthi, nangaphambi kokuba lonke uhlelo luklanywe futhi lusetshenziswe ngokugcwele. Ngaphezu kwalokho, ukubuyekezwa ngakunye okulandelayo kufanele futhi kuthathe isikhathi esincane ngangokunokwenzeka.
- Ukucwengwa okuphindaphindayo - lokhu kusho ukuthi ukuthuthukiswa ngakunye okulandelayo akufanele kuphazamise ukusebenza osekuvele kuyasebenza. Kungalesi sikhathi esivame ukuba yiphupho elibi kakhulu kumaphrojekthi amakhulu - ngokushesha noma kamuva, izinto ngazinye ziqala ukuthola ukuxhumana okuningi kangangokuthi kuba lula ukuphinda ngokuphelele i-logic ekhophi eseduze kunokwengeza insimu etafuleni elikhona. Futhi uma umangala ukuthi ukuhlaziya umthelela wokuthuthukiswa kwezinto ezikhona kungathatha isikhathi esiningi kunokuthuthukiswa ngokwazo, cishe awukakasebenzi nezinqolobane ezinkulu zedatha emabhange noma ezokuxhumana.
- Ukuzivumelanisa njalo nezimfuneko zebhizinisi ezishintshayo - Isakhiwo sento sisonke kufanele siklanywe hhayi nje kucatshangelwa ukwanda okungase kube khona, kodwa ngokulindela ukuthi isiqondiso salokhu kwandiswa okulandelayo ngeke siphuphe ngisho nasesiteji sokuklama.
Futhi yebo, ukuhlangabezana nazo zonke lezi zidingo ohlelweni olulodwa kungenzeka (Yebo, ezimweni ezithile kanye nokubhuka okuthile).
Ngezansi ngizocubungula izindlela ezimbili ezidume kakhulu zokuklama izindawo zokugcina idatha - Imodeli ye-anchor и I-Data Vault. Okusele kubakaki kunezindlela ezinhle kakhulu njengokuthi, ngokwesibonelo, i-EAV, i-6NF (ngesimo sayo esihlanzekile) nakho konke okuhlobene nezixazululo ze-NoSQL - hhayi ngoba zimbi ngandlela thize, futhi hhayi ngoba kulokhu isihloko singasongela ukuthola. umthamo we-disser emaphakathi. Ukuthi nje konke lokhu kuhlobene nezixazululo zekilasi elihluke kancane - kungaba ngamasu ongawasebenzisa ezimeni ezithile, kungakhathaliseki ukuthi iphrojekthi yakho iphelele injani (njenge-EAV), noma kwamanye amapharadigm okugcinwa kolwazi emhlabeni jikelele (njengemininingwane egciniwe yamagrafu. nezinye izinketho ze-NoSQL).
Izinkinga zendlela "yakudala" nezisombululo zazo ezindleleni eziguquguqukayo
Ngendlela "yakudala" ngiqonde inkanyezi enhle endala (kungakhathalekile ukuthi kwenziwa kanjani izendlalelo ezingaphansi, sengathi abalandeli be-Kimball, Inmon kanye ne-CDM bangixolele).
1. Ukuqina okuqinile kokuxhumana
Le modeli isuselwe ekwahlukaniseni okucacile kwedatha ibe Ubukhulu и amaqiniso. Futhi lokhu, damn it, kunengqondo - emva kwakho konke, ukuhlaziywa kwedatha eningini elikhulu lamacala kwehla ekuhlaziyweni kwezinkomba ezithile zezinombolo (amaqiniso) ezigabeni ezithile (ubukhulu).
Kulokhu, ukuxhumana phakathi kwezinto kusungulwa ngendlela yobudlelwano phakathi kwamatafula kusetshenziswa ukhiye wangaphandle. Lokhu kubonakala kungokwemvelo, kodwa ngokushesha kuholela emkhawulweni wokuqala wokuguquguquka - incazelo eqinile yekhadinali yokuxhumana.
Lokhu kusho ukuthi esigabeni sokuklama ithebula, kufanele unqume ngokunembile kupheya ngayinye yezinto ezihlobene ukuthi zingahlobana njengokuningi kuya kokuningi, noma kuphela oku-1 kuya kweziningi, kanye nokuthi “ngakuphi”. Lokhu kunquma ngokuqondile ukuthi yiliphi ithebula elizoba nokhiye oyinhloko nokuthi iliphi elizoba nokhiye wangaphandle. Ukushintsha lesi simo sengqondo lapho izidingo ezintsha zamukelwa cishe kuzoholela ekusebenzeni kabusha kwesisekelo.
Isibonelo, lapho uklama into "yerisidi yemali", wena, uthembele ezifungweni zomnyango wezokuthengisa, wabeka ukuthi kungenzeka ukwenza okuthile. iphromoshini eyodwa yezikhundla ezimbalwa zamasheke (kodwa hhayi ngokuphambene):
Futhi ngemva kwesikhathi esithile, ozakwabo bethula isu elisha lokuthengisa lapho bengenza khona isikhundla esifanayo ukukhushulwa eziningana ngesikhathi esisodwa. Futhi manje udinga ukuguqula amatafula ngokuhlukanisa ubudlelwano ube into ehlukile.
(Zonke izinto ezitholiwe lapho isheke lephromoshini lihlanganiswa khona manje nazo zidinga ukuthuthukiswa).
Ubudlelwano ku-Data Vault kanye Nemodeli Yehange
Ukugwema lesi simo kube lula kakhulu: akudingeki uthembele umnyango wezokuthengisa ukwenza lokhu. konke ukuxhumana kuqala kugcinwa kumathebula ahlukene futhi ikucubungule njengokuningi kuya kokuningi.
Le ndlela yaphakanyiswa UDan Linstedt njengengxenye yepharadigm I-Data Vault futhi isekelwe ngokugcwele Lars Rönnbäck в Imodeli Yehange.
Ngenxa yalokho, sithola isici sokuqala esihlukile sezindlela eziguquguqukayo:
Ubudlelwano phakathi kwezinto abugcinwa kuzibaluli zamabhizinisi angumzali, kodwa kuwuhlobo oluhlukile lwento.
В I-Data Vault amathebula okuxhumanisa anjalo abizwa isixhumanisi, futhi ngaphakathi Imodeli Yehange - tie. Ekuboneni kokuqala, zifana kakhulu, nakuba ukungezwani kwazo kungagcini ngegama (okuzoxoxwa ngakho ngezansi). Kuzo zombili izakhiwo, amathebula esixhumanisi angaxhuma noma iyiphi inombolo yezinhlangano (akunjalo 2).
Lokhu kuphindaphinda, ekuboneni kuqala, kunikeza ukuguquguquka okubalulekile kokulungiswa. Isakhiwo esinjalo sibekezelela hhayi kuphela izinguquko ekhadini lezixhumanisi ezikhona, kodwa futhi nokwengezwa okusha - uma manje isikhundla sesheke sinesixhumanisi ku-cashier oyivulile, ukubukeka kwesixhumanisi esinjalo kuzomane kube. isengezo phezu kwamathebula akhona ngaphandle kokuthinta noma yiziphi izinto ezikhona nezinqubo.
2. Ukuphindaphinda idatha
Inkinga yesibili exazululwe yizakhiwo eziguquguqukayo ayibonakali kahle futhi ingokwemvelo kwasekuqaleni. Izilinganiso zohlobo lwe-SCD2 (ishintsha kancane izilinganiso zohlobo lwesibili), nakuba kungezona kuphela.
Endlini yokugcina impahla yakudala, ubukhulu ngokuvamile buyithebula eliqukethe ukhiye obambele esikhundleni (njenge-PK) kanye nesethi yokhiye bebhizinisi nezibaluli kumakholomu ahlukene.
Uma ubukhulu busekela ukuguqulwa, imingcele yokufaneleka yenguqulo yengezwa kusethi evamile yezinkambu, futhi izinguqulo ezimbalwa zivela endaweni yokugcina umugqa owodwa emthonjeni (owodwa ngoshintsho ngalunye kuzibaluli zenguqulo).
Uma ubukhulu buqukethe okungenani isibaluli senguqulo esisodwa esishintsha njalo, inani lezinguqulo zobukhulu obunjalo lizomangalisa (ngisho noma izibaluli ezisele zingaguqulelwanga noma zingalokothi zishintshe), futhi uma kukhona izibaluli ezinjalo ezimbalwa, inani lezinguqulo lingashintsha. zikhule ngokukhula kusuka enanini labo. Lobu bukhulu bungathatha inani elibalulekile lesikhala sediski, nakuba idatha eningi eyigcinayo imane iyimpinda yamanani esibaluli angaguquleki asuka kweminye imigqa.
Ngesikhathi esifanayo, ibuye isetshenziswe kaningi i-denormalization - ezinye izibaluli zigcinwa ngamabomu njengenani, futhi hhayi njengesixhumanisi sencwadi yereferensi noma obunye ubukhulu. Le ndlela isheshisa ukufinyelela kwedatha, inciphisa inani lokujoyina lapho ufinyelela ubukhulu.
Ngokuvamile lokhu kuholela ulwazi olufanayo lugcinwa ngesikhathi esisodwa ezindaweni eziningana. Isibonelo, ulwazi olumayelana nesifunda sokuhlala kanye nesigaba sekhasimende lungagcinwa ngesikhathi esisodwa kubukhulu "beklayenti" kanye namaqiniso "Okuthengiwa", "Ukulethwa" kanye "Nezingcingo Zesikhungo Socingo", kanye "naku-Client - Umphathi Weklayenti. ” ithebula lesixhumanisi.
Ngokuvamile, okuchazwe ngenhla kusebenza ezilinganisweni ezivamile (ezingezona inguqulo), kodwa kwezinguqulo zingase zibe nesilinganiso esihlukile: ukubonakala kwenguqulo entsha yento (ikakhulukazi uma sibheka emuva) akuholeli kuphela ekubuyekezweni kwakho konke okuhlobene. amathebula, kodwa ekubukekeni okudabukisayo kwezinguqulo ezintsha zezinto ezihlobene - lapho iThebula 1 lisetshenziselwa ukwakha iThebula 2, futhi iThebula lesi-2 lisetshenziselwa ukwakha iThebula lesi-3, njll. Ngisho noma kungekho nesisodwa isici seThebula 1 esihilelekile ekwakhiweni kweThebula 3 (kanye nezinye izici zeThebula 2 ezitholwe kweminye imithombo zihilelekile), ukuguqulwa kwalokhu kwakhiwa kuzoholela okungenani ekuthuthukisweni okwengeziwe, futhi ngokwesilinganiso kuye kokwengeziwe. izinguqulo kuThebula 3. elingahlanganise lutho nalo nhlobo, futhi liqhubeke phansi iketango.
3. Ubunkimbinkimbi obungaqondile bokusebenza kabusha
Ngesikhathi esifanayo, isitolo sangaphambili ngasinye esisha esakhelwe phezu kwesinye sinyusa inani lezindawo lapho idatha "ingahlukana khona" lapho kwenziwa izinguquko ku-ETL. Lokhu, futhi, kuholela ekwandeni kobunzima (kanye nesikhathi) sokubuyekeza ngakunye okulandelayo.
Uma okungenhla kuchaza amasistimu anezinqubo ze-ETL ezingavamile ukuguqulwa, ungaphila ku-paradigm enjalo - udinga nje ukwenza isiqiniseko sokuthi ukuguqulwa okusha kwenziwa kahle kuzo zonke izinto ezihlobene. Uma ukubuyekezwa kwenzeka kaningi, amathuba "okulahleka" ngephutha ukuxhumana okuningana anda kakhulu.
Uma, ngaphezu kwalokho, sicabangela ukuthi i-ETL “yenguqulo” iyinkimbinkimbi kakhulu kuneyodwa “engeyona inguqulo,” kuba nzima kakhulu ukugwema amaphutha lapho uvamise ukubuyekeza sonke lesi sikhungo.
Ukugcina izinto nezibaluli ku-Data Vault kanye Nemodeli Yehange
Indlela ephakanyiswe ababhali bezakhiwo eziguquguqukayo ingakhiwa ngale ndlela elandelayo:
Kuyadingeka ukwehlukanisa ukuthi yiziphi izinguquko kulokho okuhlala kufana. Okusho ukuthi, okhiye besitolo ngokwehlukana nezibaluli.
Nokho, umuntu akufanele adideke akuhumushiwe isici nge engashintshile: eyokuqala ayiwugcini umlando wezinguquko zayo, kodwa ingashintsha (ngokwesibonelo, lapho ulungisa iphutha lokufaka noma ithola idatha entsha); eyesibili ayishintshi neze.
Amaphuzu okubuka ayahluka ngokuthi yini ngempela engabhekwa njengengaguquleki ku-Data Vault kanye ne-Anchor Model.
Ngokombono wezakhiwo I-Data Vault, kungabhekwa njengokungashintshiwe isethi yonke yezihluthulelo - yemvelo (i-TIN yenhlangano, ikhodi yomkhiqizo ohlelweni lomthombo, njll.) kanye ne-surrogate. Kulokhu, izimfanelo ezisele zingahlukaniswa ngamaqembu ngokusho komthombo kanye/noma imvamisa yezinguquko kanye Gcina itafula elihlukile leqembu ngalinye ngesethi ezimele yezinguqulo.
Ku-paradigm Imodeli Yehange kubhekwa njengokungashintshiwe ukhiye we-surrogate kuphela ingqikithi. Konke okunye (okuhlanganisa okhiye bemvelo) kumane kuyisimo esikhethekile sezimfanelo zakho. Lapho zonke izibaluli zizimele zodwa ngokuzenzakalelayo, ngakho ngemfanelo ngayinye a itafula elihlukene.
В I-Data Vault amathebula aqukethe okhiye bebhizinisi abizwa Hubami. Amahabhu ahlala equkethe isethi yezinkambu ezingashintshi:
- Okhiye Benhlangano Yemvelo
- Ukhiye we-surrogate
- Xhuma emthonjeni
- Rekhoda isikhathi sokungeza
Okuthunyelwe ku-Hubs ungalokothi ushintshe futhi awunazo izinguqulo. Ngaphandle, amahabhu afana kakhulu namathebula ohlobo lwemephu ye-ID asetshenziswa kwamanye amasistimu ukukhiqiza abanye abantu, nokho-ke, kuyanconywa ukuthi kusetshenziswe i-hashi evela kusethi yokhiye bebhizinisi njengababambeli ku-Data Vault. Le ndlela yenza kube lula ukulayisha ubudlelwano nezibaluli ezivela emithonjeni (asikho isidingo sokujoyina ihabhu ukuze uthole umuntu obambele, vele ubale i-hashi yokhiye wemvelo), kodwa kungabangela ezinye izinkinga (ezihlobene, isibonelo, ukushayisana, icala kanye nokungaphrinteki. izinhlamvu kokhiye bezintambo, njll. .p.), ngakho-ke ayamukelwa ngokuvamile.
Zonke ezinye izimfanelo zebhizinisi zigcinwa kumathebula akhethekile abizwa ngokuthi Amasathelayithi. Ihabhu elilodwa lingaba namasathelayithi amaningana agcina amasethi ahlukene ezibaluli.
Ukusatshalaliswa kwezimfanelo phakathi kwamasathelayithi kwenzeka ngokwesimiso ushintsho oluhlangene - ku-satellite eyodwa izimfanelo ezingaguquki zingagcinwa (isibonelo, usuku lokuzalwa kanye ne-SNILS yomuntu ngamunye), kwenye - okungavamile ukushintsha izinguqulo (isibonelo, isibongo nenombolo yepasipoti), kwesithathu - ezishintsha njalo (isibonelo, ikheli lokulethwa, isigaba, idethi ye-oda lokugcina, njll.). Kulokhu, ukuguqulwa kwenziwa ngezinga lamasathelayithi angawodwana, hhayi inhlangano iyonke, ngakho-ke kuyancomeka ukusabalalisa izimfanelo ukuze ukuphambana kwezinguqulo ngaphakathi kwesathelayithi eyodwa kube kuncane (okwehlisa inani eliphelele lezinguqulo ezigciniwe. ).
Futhi, ukuze kuthuthukiswe inqubo yokulayisha idatha, izibaluli ezitholwe emithonjeni ehlukahlukene zivame ukufakwa kumasathelayithi angawodwana.
Amasathelayithi axhumana nehabhu nge ukhiye wangaphandle (okuhambisana nekhadinalithi engu-1 kuya kwabaningi). Lokhu kusho ukuthi amanani ezibaluli eziningi (isibonelo, izinombolo zocingo zokuxhumana eziningi zeklayenti elilodwa) zisekelwa yile zakhiwo "ezizenzakalelayo".
В Imodeli Yehange amatafula agcina okhiye abizwa Amahange. Futhi bagcina:
- Okhiye be-surrogate kuphela
- Xhuma emthonjeni
- Rekhoda isikhathi sokungeza
Kucatshangelwa okhiye bemvelo kusukela ekubukeni kwe-Anchor Model izimfanelo ezejwayelekile. Le nketho ingase ibonakale inzima kakhulu ukuyiqonda, kodwa inikeza ububanzi obuningi bokukhomba into.
Isibonelo, uma idatha emayelana nebhizinisi elifanayo ingavela kumasistimu ahlukene, ngayinye esebenzisa ukhiye wayo wemvelo. Ku-Data Vault, lokhu kungaholela ezakhiweni ezinzima kakhulu zamahabhu amaningana (eyodwa ngomthombo + inguqulo eyinhloko ehlanganisayo), kuyilapho kumodeli ye-Anchor, ukhiye wemvelo womthombo ngamunye uwela esibalulweni sawo futhi ungasetshenziswa uma ulayisha ngaphandle kosizo. bonke abanye.
Kodwa kukhona nephuzu elilodwa elicashile lapha: uma izibaluli ezivela ezinhlelweni ezihlukene zihlanganiswa enhlanganweni eyodwa, kungenzeka ukuthi kukhona Imithetho ye-"gluing", isistimu okufanele iqonde ngayo ukuthi amarekhodi avela emithonjeni ehlukene ahambisana nesenzakalo esisodwa sebhizinisi.
В I-Data Vault le mithetho cishe iyonquma ukwakheka "ihabhu yokutholwa" yebhizinisi eliyinhloko futhi awanathonya nganoma iyiphi indlela Amahabhu agcina okhiye bomthombo wemvelo kanye nezici zabo zangempela. Uma ngesikhathi esithile imithetho yokuhlanganisa ishintsha (noma izibaluli okwenziwa ngazo zibuyekezwa), kuzokwanela ukufometha kabusha amahabhu okungena esikhundleni.
В Imodeli ye-anchor ibhizinisi elinjalo cishe lizogcinwa kulo ihange kuphela. Lokhu kusho ukuthi zonke izici, kungakhathaliseki ukuthi zivela kumuphi umthombo, zizoboshelwa kumuntu obambele omunye. Ukuhlukanisa amarekhodi ahlanganiswe ngephutha futhi, ngokuvamile, ukuqapha ukuhlobana kokuhlanganiswa ohlelweni olunjalo kungaba nzima kakhulu, ikakhulukazi uma imithetho iyinkimbinkimbi futhi ishintsha njalo, futhi isici esifanayo singatholakala emithonjeni ehlukene (yize ngokuqinisekile kungenzeka, njengoba inguqulo ngayinye yesibaluli igcina isixhumanisi emthonjeni wayo).
Kunoma ikuphi, uma isistimu yakho kufanele iqalise ukusebenza ukukhishwa, ukuhlanganisa amarekhodi nezinye izici ze-MDM, kufanelekile ukunaka ngokukhethekile izici zokugcina okhiye bemvelo ezindleleni ezigugile. Kungenzeka ukuthi umklamo omkhulu we-Data Vault uzovikeleka ngokuphazima kweso mayelana namaphutha okuhlanganisa.
Imodeli ye-anchor futhi inikeza uhlobo lwento eyengeziwe ebizwa ngokuthi Ifindo kukhethekile uhlobo oluwohlokayo lwehange, okungaqukatha isibaluli esisodwa kuphela. Amanodi kufanele asetshenziselwe ukugcina uhlu lwemibhalo oluyisicaba (isibonelo, ubulili, isimo somshado, isigaba senkonzo yamakhasimende, njll.). Ngokungafani neAnchor, Ifindo ayinawo amathebula esibaluli ahlobene, futhi okuwukuphela kwesibaluli sayo (igama) lihlala ligcinwe kuthebula elifanayo nokhiye. Ama-Node axhunywe kuma-Anchors ngamatafula okubopha (Tie) ngendlela efanayo nama-Anchors axhunywe komunye nomunye.
Awukho umbono ocacile mayelana nokusetshenziswa kwamaNodes. Ngokwesibonelo, , okhuthaza ngenkuthalo ukusetshenziswa kwe-Anchor Model eRussia, ukholelwa (hhayi ngokungafanele) ukuthi akukho neyodwa incwadi yereferensi okungashiwo ngokuqiniseka ukuthi njalo izoba yi-static futhi ibe neleveli eyodwa, ngakho-ke kungcono ukusebenzisa ngokushesha i-Anchor egcwele kuzo zonke izinto.
Omunye umehluko obalulekile phakathi kwe-Data Vault kanye nemodeli ye-Anchor ukutholakala izimfanelo zokuxhuma:
В I-Data Vault Izixhumanisi ziyizinto ezifanayo ezigcwele ngokugcwele njengamaHub, futhi zingaba nazo izimfanelo siqu. I Imodeli ye-anchor Izixhumanisi zisetshenziswa kuphela ukuxhuma Izikhonkwane kanye abakwazi ukuba nezabo izici. Lo mehluko uphumela ezindleleni zokumodela ezihluke kakhulu amaqiniso, okuzoxoxwa ngakho ngokuqhubekayo.
Ukugcinwa kweqiniso
Ngaphambi kwalokhu, sikhulume kakhulu ngokulinganiswa kokulinganisa. Amaqiniso awacacile kancane.
В I-Data Vault into ejwayelekile yokugcina amaqiniso ngu Isixhumanisi, okufakwe kumasathelayithi izinkomba zangempela.
Le ndlela ibonakala inembile. Inikeza ukufinyelela okulula ezinkomba ezihlaziywe futhi ngokuvamile ifana nethebula leqiniso lendabuko (izinkomba kuphela azigcinwa etafuleni ngokwalo, kodwa kwelithi "elingumakhelwane"). Kepha kukhona nezingibe: enye yezinguquko ezijwayelekile zemodeli - ukunwetshwa kokhiye weqiniso - kudinga ingeza ukhiye omusha wangaphandle kokuthi Xhumanisa. Futhi lokhu, futhi, "kwephula" i-modularity futhi kungabangela isidingo sokuguqulwa kwezinye izinto.
В Imodeli ye-anchor Uxhumano alukwazi ukuba nezimfanelo zalo, ngakho le ndlela ngeke isebenze - zonke izimfanelo nezinkomba kufanele zixhunywe kuhange elithile. Isiphetho esivela kulokhu silula - Iqiniso ngalinye nalo lidinga ihange lalo. Kokunye esijwayele ukukubona njengamaqiniso, lokhu kungase kubukeke kungokwemvelo - isibonelo, iqiniso lokuthenga lingancishiswa ngokuphelele libe into "i-oda" noma "iresidi", ukuvakashela isayithi kuseshini, njll. Kodwa kukhona namaqiniso okungekho lula ukuthola "into yenkampani" enjalo yemvelo - isibonelo, izinsalela zezimpahla ezindaweni zokugcina impahla ekuqaleni kosuku ngalunye.
Ngakho-ke, izinkinga nge-modularity lapho unweba ukhiye weqiniso kumodeli ye-Anchor aziveli (kwanele ukumane wengeze ubudlelwano obusha ku-Anchor ehambisanayo), kodwa ukuklama imodeli yokubonisa amaqiniso akulula kangako; Izikhonkwane "zokwenziwa" zingase zivele. ezibonisa imodeli yento yebhizinisi ngendlela engacacile.
Ukuvumelana nezimo kufinyelelwa kanjani
Ukwakhiwa okuphumela kuzo zombili izimo kuqukethe amatafula ngokuphawulekayo ngaphezulukunokulinganisa kwendabuko. Kodwa kungase kuthathe isikhala sediski sincane kakhulu ngesethi efanayo yezibaluli zenguqulo njengobukhulu bendabuko. Ngokwemvelo, akukho umlingo lapha - konke kumayelana nokujwayelekile. Ngokusabalalisa izibaluli kuwo wonke amaSathelayithi (ku-Data Vault) noma amathebula angawodwana (Imodeli Yesikhonkwane), sehlisa (noma sisuse ngokuphelele) ukuphindwaphindwa kwamanani ezinye izimfanelo lapho ushintsha ezinye.
Ukuze I-Data Vault ukuwina kuzoncika ekusabalaliseni kwezimfanelo phakathi kwamaSathelayithi, futhi Imodeli ye-anchor — icishe ilingane ngokuqondile nenani elijwayelekile lezinguqulo ngento ngayinye yokulinganisa.
Kodwa-ke, ukonga isikhala kubalulekile, kodwa hhayi okuyinhloko, inzuzo yokugcina izimfanelo ngokwehlukana. Kanye nesitoreji esihlukile sobudlelwano, le ndlela yenza isitolo design Modular. Lokhu kusho ukuthi ukungeza kokubili izibaluli ezingazodwana kanye nezindawo zezifundo ezintsha eziphelele kumodeli enjalo kubukeka kanje superstructure phezu kwesethi ekhona yezinto ngaphandle kokuzishintsha. Futhi yilokhu kanye okwenza izindlela ezichaziwe zivumelane nezimo.
Lokhu kuphinde kufane noguquko olusuka ekukhiqizeni izingcezu luye ekukhiqizeni ngobuningi - uma ngendlela yendabuko ithebula ngalinye lemodeli lihlukile futhi lidinga ukunakwa okukhethekile, khona-ke ezindleleni eziguquguqukayo selivele liyisethi "yezingxenye" ezijwayelekile. Ngakolunye uhlangothi, kunamathebula amaningi, futhi izinqubo zokulayisha kanye nokubuyisa idatha kufanele zibukeke ziyinkimbinkimbi kakhulu. Ngakolunye uhlangothi, baba ejwayelekile. Okusho ukuthi kungenzeka kube khona okuzenzakalelayo kanye nemethadatha. Umbuzo othi "sizoyibeka kanjani?", impendulo engase ithathe ingxenye ebalulekile yomsebenzi wokuklama ukuthuthukiswa, manje awufanelekile (kanye nombuzo mayelana nomthelela wokushintsha imodeli ezinqubweni zokusebenza. ).
Lokhu akusho ukuthi abahlaziyi abadingeki ohlelweni olunjalo nhlobo - othile usadinga ukusebenzisa isethi yezinto ezinezimfanelo futhi athole ukuthi uzokulayisha kuphi futhi kanjani konke. Kodwa inani lomsebenzi, kanye nokwenzeka kanye nezindleko zephutha, ziyancipha kakhulu. Kokubili esigabeni sokuhlaziya nangesikhathi sokuthuthukiswa kwe-ETL, okuthi engxenyeni ebalulekile ingancishiswa ibe imethadatha yokuhlela.
Isayidi elimnyama
Konke lokhu okungenhla kwenza zombili izindlela zivumelane nezimo ngempela, zithuthuke kakhulu kwezobuchwepheshe futhi zifanele ukuthuthuka okuphindaphindiwe. Yiqiniso, kukhona futhi "umgqomo emafutheni", engicabanga ukuthi usuvele uqagele ngawo.
Ukuwohloka kwedatha, okungaphansi kwe-modularity yezakhiwo eziguquguqukayo, kuholela ekwandeni kwenani lamatafula futhi, ngokufanele, phezulu ukujoyina lapho kwenziwa isampula. Ukuze uthole kalula zonke izici zobukhulu, esitolo sakudala ukukhetha okukodwa kwanele, kodwa ukwakheka okuguquguqukayo kuzodinga uchungechunge lonke lokujoyina. Ngaphezu kwalokho, uma konke lokhu kujoyina kwemibiko kungabhalwa kusenesikhathi, abahlaziyi abajwayele ukubhala i-SQL ngesandla bazohlupheka kabili.
Kunamaqiniso amaningana enza lesi simo sibe lula:
Lapho usebenza ngobukhulu obukhulu, zonke izici zayo cishe azikaze zisetshenziswe ngasikhathi sinye. Lokhu kusho ukuthi kungase kube nokujoyina okumbalwa kunalokho okubonakala uma uqala ukuyibuka kumodeli. I-Data Vault ingase futhi icabangele imvamisa elindelekile yokwabelana lapho kwabiwa izimfanelo kumasathelayithi. Ngesikhathi esifanayo, ama-Hub noma ama-Anchors ngokwawo ayadingeka ngokuyinhloko ekukhiqizeni nasekudwebeni ama-surrogates esigabeni sokulayisha futhi awavamile ukusetshenziswa emibuzweni (lokhu kuyiqiniso ikakhulukazi kuma-Anchors).
Wonke amajoyina angokhiye. Ukwengeza, indlela "ecindezelwe" kakhulu yokugcina idatha inciphisa ingaphezulu lamathebula okuskena lapho idingeka khona (isibonelo, lapho uhlungwa ngevelu yesibaluli). Lokhu kungaholela eqinisweni lokuthi ukusampula kusizindalwazi esijwayelekile esinenqwaba yokujoyina kuzoshesha kakhulu kunokuskena ubukhulu obubodwa obusindayo ngezinguqulo eziningi umugqa ngamunye.
Ngokwesibonelo, lapha I-athikili iqukethe isivivinyo sokuqhathanisa esinemininingwane sokusebenza kwemodeli ye-Anchor ngesampula evela kuthebula elilodwa.
Okuningi kuncike enjinini. Izinkundla eziningi zesimanje zinezindlela zokuthuthukisa ukujoyina kwangaphakathi. Isibonelo, i-MS SQL ne-Oracle “zingeqa” ukujoyina amathebula uma idatha yazo ingasetshenziswa noma kuphi ngaphandle kwamanye ama-joyinti futhi akuthinti ukukhetha kokugcina (ukususwa kwethebula/joyina), kanye ne-MPP Vertica. , ibonakale iyinjini esezingeni eliphezulu kakhulu ye-Anchor Model, uma kubhekwa ukwenziwa kahle kwecebo lombuzo. Ngakolunye uhlangothi, ukugcina i-Anchor Model, isibonelo, ku-Click House, enokusekelwa okulinganiselwe kokujoyina, akubukeki njengombono omuhle kakhulu.
Ngaphezu kwalokho, kuzo zombili izakhiwo zikhona ukunyakaza okukhethekile, okwenza ukufinyelela kwedatha kube lula (kokubili ngokombono wokusebenza kombuzo nakubasebenzisi bokugcina). Ngokwesibonelo, Amathebula e-Point-In-Time ku-Data Vault noma imisebenzi ekhethekile yetafula kumodeli ye-Anchor.
Inani
Ingqikithi eyinhloko yezakhiwo ezicatshangelwayo eziguquguqukayo yi-modularity "yomklamo" wabo.
Lesi sakhiwo esivumela:
- Ngemuva kokulungiselela okuthile kokuqala okuhlobene nokuthunyelwa kwemethadatha nokubhala ama-algorithms ayisisekelo e-ETL, ngokushesha unikeze ikhasimende umphumela wokuqala ngendlela yemibiko embalwa equkethe idatha evela kumthombo wezinto ezimbalwa. Akudingekile ukucabanga ngokuphelele (ngisho sezingeni eliphezulu) yonke imodeli yento.
- Imodeli yedatha ingaqala ukusebenza (futhi ibe wusizo) ngezinto ezi-2-3 kuphela, bese-ke khula kancane kancane (Mayelana nemodeli ye-Anchor u-Nikolai ukuqhathanisa okuhle ne-mycelium).
- Ukuthuthukiswa okuningi, okuhlanganisa ukunweba indawo yesihloko nokwengeza imithombo emisha ayithinti ukusebenza okukhona futhi ayibeki engcupheni yokuphula into esivele iyasebenza.
- Ngenxa yokubola ibe yizici ezijwayelekile, izinqubo ze-ETL ezinhlelweni ezinjalo zibukeka zifana, ukubhala kwazo kuziboleka ku-algorithmization futhi, ekugcineni, okuzenzakalelayo.
Intengo yalokhu kuvumelana nezimo ukusebenza. Lokhu akusho ukuthi akunakwenzeka ukufeza ukusebenza okwamukelekayo kumamodeli anjalo. Kaningi kunalokho, ungase udinge umzamo owengeziwe nokunaka imininingwane ukuze ufinyelele amamethrikhi owafunayo.
Izicelo
Izinhlobo zebhizinisi I-Data Vault
Olunye ulwazi mayelana ne-Data Vault:
Izinhlobo zebhizinisi Imodeli Yehange
Imininingwane eminingi nge-Anchor Model:
Ithebula elifingqiwe elinezici ezivamile kanye nomehluko wezindlela ezicatshangelwayo:
Source: www.habr.com