Kubonakala sengathi inkambu yokukhangisa ku-inthanethi kufanele ithuthuke ngokobuchwepheshe futhi izenzekele ngangokunokwenzeka. Yiqiniso, ngoba ama-giants nochwepheshe emkhakheni wabo njenge-Yandex, Mail.Ru, Google ne-Facebook basebenza lapho. Kodwa, njengoba kwavela, akukho mkhawulo ekupheleleni futhi kukhona njalo okuthile okuzenzakalelayo.
Iqembu lezokuxhumana I-Dentsu Aegis Network Russia ingumdlali omkhulu kunabo bonke emakethe yokukhangisa yedijithali futhi itshala imali ngenkuthalo kubuchwepheshe, izama ukuthuthukisa nokwenza ngokuzenzakalela izinqubo zayo zebhizinisi. Enye yezinkinga ezingaxazululiwe zemakethe yokukhangisa ku-inthanethi umsebenzi wokuqoqa izibalo zemikhankaso yokukhangisa ezisekelweni ze-inthanethi ezihlukene. Isixazululo sale nkinga ekugcineni siholele ekwakhiweni komkhiqizo D1.Idijithali (funda njenge-DiVan), intuthuko esifuna ukukhuluma ngayo.
Kungani?
1. Ngesikhathi sokuqala kwephrojekthi, kwakungekho nowodwa umkhiqizo owenziwe ngomumo emakethe owaxazulula inkinga yokuzenzakalelayo ukuqoqwa kwezibalo emikhankasweni yokukhangisa. Lokhu kusho ukuthi akekho ngaphandle kwethu oyokwanelisa izidingo zethu.
Amasevisi afana ne-Improvado, i-Roistat, i-Supermetrics, i-SegmentStream inikeza ukuhlanganiswa nezinkundla, izinkundla zokuxhumana kanye ne-Google Anaalitycs, futhi yenza kube nokwenzeka ukwakha amadeshibhodi okuhlaziya ukuze kuhlaziywe futhi kulawulwe imikhankaso yokukhangisa. Ngaphambi kokuthi siqale ukuthuthukisa umkhiqizo wethu, sizamile ukusebenzisa amanye alawa masistimu ukuze siqoqe idatha kumasayithi, kodwa, ngeshwa, awakwazanga ukuxazulula izinkinga zethu.
Inkinga enkulu yayiwukuthi imikhiqizo ehloliwe yayisekelwe emithonjeni yedatha, ibonisa izibalo zokubeka ngesayithi, futhi ayizange inikeze amandla okuhlanganisa izibalo zemikhankaso yokukhangisa. Le ndlela ayizange isivumele ukuthi sibone izibalo ezivela kumasayithi ahlukene endaweni eyodwa futhi sihlaziye isimo somkhankaso wonkana.
Esinye isici ukuthi ezinyathelweni zokuqala imikhiqizo yayihloswe emakethe yaseNtshonalanga futhi ayizange isekele ukuhlanganiswa nezingosi zaseRussia. Futhi kulawo masayithi ukuhlanganiswa okwenziwa ngawo, wonke amamethrikhi adingekayo awazange alandwe ngemininingwane eyanele ngaso sonke isikhathi, futhi ukuhlanganiswa kwakungelula ngaso sonke isikhathi futhi kusobala, ikakhulukazi lapho kudingekile ukuthola okuthile okungekho kusixhumi esibonakalayo sesistimu.
Ngokuvamile, sinqume ukungazivumelani nemikhiqizo yezinkampani zangaphandle, kodwa saqala ukuthuthukisa eyethu...
2. Imakethe yokukhangisa ku-inthanethi iyakhula unyaka nonyaka, futhi ngo-2018, ngokwezabelomali zokukhangisa, yadlula imakethe yokukhangisa ye-TV enkulu ngokwesiko. Ngakho kukhona isikali.
3. Ngokungafani nemakethe yokukhangisa ye-TV, lapho ukuthengiswa kokukhangisa kwezentengiso kulawulwa yedwa, kunenqwaba yabanikazi abangabodwana bohlu lokukhangisa losayizi abahlukahlukene abasebenza ku-inthanethi ngama-akhawunti abo okukhangisa. Njengoba umkhankaso wokukhangisa, njengomthetho, ugijima kumasayithi amaningana ngesikhathi esisodwa, ukuze uqonde isimo somkhankaso wokukhangisa, kuyadingeka ukuqoqa imibiko evela kuzo zonke izingosi futhi uyihlanganise ibe umbiko owodwa omkhulu ozobonisa isithombe sonke. Lokhu kusho ukuthi akhona amandla okuthuthukisa.
4. Kithina kubonakala sengathi abanikazi bempahla yokukhangisa ku-inthanethi sebevele benengqalasizinda yokuqoqa izibalo nokuziveza kuma-akhawunti okukhangisa, futhi bazokwazi ukuhlinzeka nge-API yale datha. Lokhu kusho ukuthi ngokobuchwepheshe kungenzeka ukukusebenzisa. Ake sithi zisuka nje kwavela ukuthi akulula kangako.
Ngokuvamile, zonke izimfuneko zokuqaliswa kwephrojekthi zazisobala kithi, futhi sagijima ukuze senze iphrojekthi iphile ...
Uhlelo olukhulu
Okokuqala, sakhe umbono wesistimu efanelekile:
- Imikhankaso yokukhangisa evela kusistimu yebhizinisi ye-1C kufanele ilayishwe kuyo ngokuzenzakalelayo namagama ayo, izikhathi, amabhajethi kanye nokubekwa ezinkundleni ezihlukahlukene.
- Ekubekweni ngakunye phakathi komkhankaso wokukhangisa, zonke izibalo ezingaba khona kufanele zilandwe ngokuzenzakalelayo kumasayithi lapho ukubekwa kwenzeka khona, njengenombolo yokuvelayo, ukuchofoza, ukubuka, njll.
- Eminye imikhankaso yokukhangisa ilandelelwa kusetshenziswa ukuqapha kwenkampani yangaphandle ngalokho okubizwa ngamasistimu okukhangisa afana ne-Adriver, i-Weborama, i-DCM, njll. Kukhona futhi imitha ye-Intanethi yezimboni eRussia - inkampani yeMediascope. Ngokohlelo lwethu, idatha evela ekuqashweni okuzimele kanye nezimboni kufanele futhi ilayishwe ngokuzenzakalelayo emikhankasweni yokukhangisa ehambisanayo.
- Iningi lemikhankaso yokukhangisa ku-inthanethi ihloselwe izenzo ezithile eziqondiwe (ukuthenga, ukushayela, ukubhalisela idrayivu yokuhlola, njll.), elandelelwa kusetshenziswa i-Google Analytics, kanye nezibalo zazo nazo ezibalulekile ekuqondeni isimo somkhankaso kanye kufanele kulayishwe ethuluzini lethu.
Into yokuqala eyi-lumpy
Uma kubhekwa ukuzibophezela kwethu ezimisweni eziguquguqukayo zokuthuthukiswa kwesofthiwe (ishesha, zonke izinto), sinqume ukuthi siqale sithuthukise i-MVP bese sidlulela emgomweni esihlosiwe ngokuphindaphindiwe.
Sinqume ukwakha i-MVP ngokusekelwe kumkhiqizo wethu I-DANBo (Dentsu Aegis Network Board), okuwuhlelo lwewebhu olunolwazi olujwayelekile emikhankasweni yokukhangisa yamakhasimende ethu.
Ku-MVP, iphrojekthi yenziwe yaba lula ngangokunokwenzeka mayelana nokuqaliswa. Sikhethe uhlu olulinganiselwe lwezinkundla zokuhlanganisa. Lawa bekuyizinkundla eziyinhloko, ezifana ne-Yandex.Direct, Yandex.Display, RB.Mail, MyTarget, Adwords, DBM, VK, FB, kanye nezinhlelo eziyinhloko zokukhangisa i-Adriver ne-Weborama.
Ukufinyelela izibalo kumasayithi nge-API, sisebenzise i-akhawunti eyodwa. Umphathi weqembu leklayenti obefuna ukusebenzisa ukuqoqwa okuzenzakalelayo kwezibalo emkhankasweni wokukhangisa bekumele aqale athumele ukufinyelela emikhankasweni yokukhangisa edingekayo kumasayithi ku-akhawunti yenkundla.
Okulandelayo umsebenzisi wesistimu I-DANBO kwadingeka ilayishe ifayela lefomethi ethile ohlelweni lwe-Excel, oluqukethe lonke ulwazi mayelana nokubekwa (umkhankaso wokukhangisa, inkundla, ifomethi, isikhathi sokubeka, izinkomba ezihleliwe, isabelomali, njll.) kanye nezihlonzi zemikhankaso yokukhangisa ehambisanayo amasayithi kanye nezinto zokubala ezinhlelweni zokukhangisa.
Kwakubonakala, ngokungananazi, kwethusa:
Idatha elandiwe ilondolozwe kusizindalwazi, kwase kuba amasevisi ahlukene aqoqa izihlonzi zomkhankaso kumasayithi asuka kuzo futhi alanda izibalo kuzo.
Kusayithi ngalinye, kwabhalwa isevisi ehlukile yamawindi, eyathi kanye ngosuku ibe ngaphansi kwe-akhawunti yesevisi eyodwa ku-API yesayithi futhi yalanda izibalo zama-ID omkhankaso othile. Kwenzeka okufanayo nangezinhlelo zokukhangisa.
Idatha elandiwe iboniswe kusixhumi esibonakalayo ngesimo sedeshibhodi encane yangokwezifiso:
Ngokungalindelekile kithi, i-MVP yaqala ukusebenza futhi yaqala ukulanda izibalo zamanje zemikhankaso yokukhangisa ku-inthanethi. Sisebenzise uhlelo kumakhasimende amaningana, kodwa lapho sizama ukukala, sihlangabezane nezinkinga ezinkulu:
- Inkinga enkulu kwaba inkimbinkimbi yokulungiselela idatha ukuze ilayishwe ohlelweni. Futhi, idatha yokubeka bekufanele iguqulwe ibe ifomethi eqinile ngaphambi kokulayisha. Bekudingeka ukuthi kufakwe izihlonzi zebhizinisi ezisuka kumasayithi ahlukene efayeleni lokulanda. Sibhekene neqiniso lokuthi kunzima kakhulu kubasebenzisi abangaqeqeshiwe ngokobuchwepheshe ukuchaza ukuthi bangazithola kuphi lezi zihlonzi kusayithi nokuthi zidinga ukufakwa kuphi efayeleni. Uma kucatshangelwa inani labasebenzi eminyangweni abaqhuba imikhankaso ezizindeni kanye nenzuzo, lokhu kuholele ekusekelweni okukhulu ngasohlangothini lwethu, esingeneme neze ngakho.
- Enye inkinga yayiwukuthi akuzona zonke izinkundla zokukhangisa ezazinezindlela zokudlulisela ukufinyelela emikhankasweni yokukhangisa kwamanye ama-akhawunti. Kodwa noma ngabe bekukhona indlela yokuthumela, akubona bonke abakhangisi abebezimisele ukunikeza ukufinyelela emikhankasweni yabo kuma-akhawunti ezinkampani zangaphandle.
- Isici esibalulekile kwaba ukucasuka okwavusa abasebenzisi ukuthi zonke izinkomba ezihleliwe kanye nemininingwane yokubeka asebevele beyifaka ohlelweni lwethu lokubala lwe-1C, kufanele baphinde bangene I-DANBO.
Lokhu kusinike umbono wokuthi umthombo oyinhloko wolwazi mayelana nokubekwa kufanele kube isistimu yethu ye-1C, lapho yonke idatha ifakwa khona ngokunembile nangesikhathi (iphuzu lapha ukuthi ama-invoyisi akhiqizwa ngokusekelwe kudatha ye-1C, ngakho ukufakwa okulungile kwedatha ku-1C ibalulekile kuwo wonke umuntu we-KPI). Wavela kanje umqondo omusha wesistimu...
Umqondo
Into yokuqala esinqume ukuyenza kwakuwukuhlukanisa uhlelo lokuqoqa izibalo zemikhankaso yokukhangisa ku-inthanethi zibe umkhiqizo ohlukile - D1.Idijithali.
Emcabangweni omusha, sinqume ukulayisha kuwo D1.Idijithali ulwazi lwemikhankaso yokukhangisa nokubekwa ngaphakathi kwayo ukusuka ku-1C, bese ukhipha izibalo kumasayithi namasistimu e-AdServing kuya kulokhu kubekwa. Lokhu bekufanele kwenze kube lula kakhulu ukuphila kubasebenzisi (futhi, njengenjwayelo, kwengeze umsebenzi owengeziwe kubathuthukisi) futhi kunciphise inani losekelo.
Inkinga yokuqala esihlangabezane nayo kwakungeyenhlangano futhi yayihlobene neqiniso lokuthi asikwazanga ukuthola ukhiye noma uphawu esingaqhathanisa ngalo amabhizinisi avela kumasistimu ahlukene nemikhankaso nokubekwa okuvela ku-1C. Iqiniso liwukuthi inqubo enkampanini yethu iklanywe ngendlela yokuthi imikhankaso yokukhangisa ifakwe ezinhlelweni ezahlukene ngabantu abahlukene (abahleli bemidiya, ukuthenga, njll.).
Ukuze sixazulule le nkinga, kudingeke ukuthi sisungule ukhiye we-hashi oyingqayizivele, i-DANBoID, ongaxhumanisa amabhizinisi kumasistimu ahlukene ndawonye, ββfuthi ongabonakala kalula nangokuhlukile kumasethi edatha alandiwe. Lesi sihlonzi sikhiqizwa ohlelweni lwangaphakathi lwe-1C ekubekweni ngakunye ngakunye futhi sidluliselwa emikhankasweni, ezindaweni ezibekiwe nasezibalini kuwo wonke amasayithi nakuwo wonke amasistimu e-AdServing. Ukuqalisa umkhuba wokufaka i-DANBoID kukho konke ukubekwa kwathatha isikhathi, kodwa sikwazile ukukwenza :)
Sabe sesithola ukuthi akuwona wonke amasayithi ane-API yokuqoqa ngokuzenzakalelayo izibalo, futhi nalawo ane-API, ayibuyisi yonke idatha edingekayo.
Kulesi sigaba, sinqume ukunciphisa kakhulu uhlu lwezinkundla zokuhlanganiswa futhi sigxile ezinkundleni eziyinhloko ezihilelekile emikhankasweni eminingi yokukhangisa. Lolu hlu luhlanganisa bonke abadlali abakhulu emakethe yokukhangisa (i-Google, i-Yandex, i-Mail.ru), izinkundla zokuxhumana (VK, Facebook, Twitter), izinhlelo ezinkulu ze-AdServing ne-analytics (DCM, Adriver, Weborama, Google Analytics) nezinye izinkundla.
Iningi lamasayithi esiwakhethile ane-API ehlinzeka ngamamethrikhi esiwadingayo. Ezimeni lapho ibingenayo i-API noma ingenayo idatha edingekayo, sisebenzisa imibiko ethunyelwa nsuku zonke ku-imeyili yehhovisi lethu ukuze silayishe idatha (kwezinye izinhlelo kungenzeka ukulungisa imibiko enjalo, kwezinye sivumelene ngokuthuthukiswa kwemibiko enjalo. okwethu).
Lapho sihlaziya idatha evela kumasayithi ahlukene, sithole ukuthi isigaba samabhizinisi asifani kumasistimu ahlukene. Ngaphezu kwalokho, ulwazi ludinga ukulandwa ngemininingwane ehlukene kumasistimu ahlukene.
Ukuxazulula le nkinga, umqondo we-SubDANBoID wathuthukiswa. Umqondo we-SubDANBoID ulula impela, simaka inkampane eyinhloko yomkhankaso esizeni nge-DANBoID ekhiqiziwe, futhi silayisha zonke izinhlangano ezifakwe ezidlekeni ezinezihlonzi zesayithi ezihlukile bese senza i-SubDANBoID ngokomgomo we-DANBoID + isihlonzi sezinga lokuqala. ibhizinisi elikhiqizwe + inkomba yezinga lesibili lebhizinisi elikhiqizwe +... Le ndlela yasivumela ukuthi sixhume imikhankaso yokukhangisa kumasistimu ahlukene futhi silande izibalo ezinemininingwane kuyo.
Kwadingeka futhi sixazulule inkinga yokufinyelela imikhankaso ezinkundleni ezahlukene. Njengoba sibhale ngenhla, indlela yokudlulisela ukufinyelela kumkhankaso ku-akhawunti ehlukile yobuchwepheshe ayisebenzi ngaso sonke isikhathi. Ngakho-ke, kudingeke ukuthi sakhe ingqalasizinda yokugunyazwa okuzenzakalelayo nge-OAuth sisebenzisa amathokheni nezindlela zokubuyekeza lawa mathokheni.
Kamuva esihlokweni sizozama ukuchaza ngokuningiliziwe ukwakheka kwesixazululo kanye nemininingwane yezobuchwepheshe yokuqaliswa.
I-Solution Architecture 1.0
Lapho siqala ukuqaliswa komkhiqizo omusha, saqonda ukuthi sasidinga ngokushesha ukunikeza ithuba lokuxhuma amasayithi amasha, ngakho-ke sanquma ukulandela indlela ye-microservice architecture.
Lapho siklama izakhiwo, sihlukanise izixhumi kuwo wonke amasistimu angaphandle - 1C, izinkundla zokukhangisa namasistimu okukhangisa - sibe yizinsiza ezihlukene.
Umbono oyinhloko ukuthi zonke izixhumi kumasayithi zine-API efanayo futhi zingama-adaptha aletha i-API yesayithi kusixhumi esibonakalayo esilungele thina.
Enkabeni yomkhiqizo wethu kunesicelo sewebhu, okuyi-monolith eklanywe ngendlela yokuthi ingahlakazwa kalula ibe yizinsizakalo. Lolu hlelo lokusebenza lunesibopho sokucubungula idatha elandiwe, ukuhlanganisa izibalo ezivela kumasistimu ahlukene futhi izethule kubasebenzisi besistimu.
Ukuze sixhumane phakathi kwezixhumi nohlelo lokusebenza lwewebhu, bekufanele sakhe isevisi eyengeziwe, esiyibiza ngokuthi Ummeleli Wesixhumi. Lenza imisebenzi Yokutholwa Kwesevisi kanye Nesihleli Somsebenzi. Le sevisi isebenzisa imisebenzi yokuqoqa idatha yesixhumi ngasinye njalo ebusuku. Ukubhala isendlalelo sesevisi bekulula kunokuxhuma umthengisi womlayezo, futhi kithina bekubalulekile ukuthola umphumela ngokushesha okukhulu.
Ukuze kube lula kanye nesivinini sokuthuthuka, siphinde sanquma ukuthi zonke izinsiza zizoba ama-Web API. Lokhu kwenze kwaba nokwenzeka ukuhlanganisa ngokushesha umqondo wobufakazi futhi kuqinisekiswe ukuthi umklamo wonke uyasebenza.
Umsebenzi ohlukile, oyinkimbinkimbi kwakuwukusetha ukufinyelela ukuze kuqoqwe idatha kuma-akhawunti ahlukene, lokho, njengoba sinqume, okufanele wenziwe abasebenzisi ngokusebenzisa isixhumi esibonakalayo sewebhu. Iqukethe izinyathelo ezimbili ezihlukene: okokuqala, umsebenzisi wengeza ithokheni ukuze afinyelele i-akhawunti nge-OAuth, bese elungiselela ukuqoqwa kwedatha yeklayenti kusuka ku-akhawunti ethile. Ukuthola ithokheni nge-OAuth kuyadingeka ngoba, njengoba sesivele sibhalile, akwenzeki ngaso sonke isikhathi ukunikeza ukufinyelela ku-akhawunti oyifunayo kusayithi.
Ukuze udale indlela yendawo yonke yokukhetha i-akhawunti kumasayithi, bekufanele sengeze indlela kuzixhumi ze-API ezibuyisela i-JSON Schema, enikezwa ifomu kusetshenziswa ingxenye eguquliwe ye-JSONEditor. Ngale ndlela, abasebenzisi bakwazile ukukhetha ama-akhawunti abazolanda kuwo idatha.
Ukuze sihambisane nemikhawulo yesicelo ekhona kumasayithi, sihlanganisa izicelo zezilungiselelo ngaphakathi kwethokheni eyodwa, kodwa singakwazi ukucubungula amathokheni ahlukene ngokuhambisana.
Sikhethe i-MongoDB njengendawo yokugcina idatha elayishiwe yakho kokubili uhlelo lokusebenza lwewebhu nezixhumi, okusivumele ukuthi singakhathazeki kakhulu ngesakhiwo sedatha ezigabeni zokuqala zokuthuthuka, lapho imodeli yento yohlelo lokusebenza ishintsha njalo ngemva kolunye usuku.
Ngokushesha sithole ukuthi akuyona yonke idatha elingana kahle ku-MongoDB futhi, isibonelo, kulula kakhulu ukugcina izibalo zansuku zonke kusizindalwazi esihlobene. Ngakho-ke, kuzixhumi izakhiwo zazo zedatha zifaneleka kakhulu kusizindalwazi esihlobene, siqale ukusebenzisa i-PostgreSQL noma i-MS SQL Server njengesitoreji.
Izakhiwo ezikhethiwe kanye nobuchwepheshe kusivumele ukuthi sakhe futhi sethule umkhiqizo we-D1.Digital ngokushesha uma kuqhathaniswa. Ngaphezulu kweminyaka emibili yokuthuthukiswa komkhiqizo, sithuthukise izixhumi ezingu-23 kumasayithi, sazuza ulwazi olubalulekile lokusebenza nama-API ezinkampani zangaphandle, safunda ukugwema izingibe zamasayithi ahlukene, ngayinye eyayineyakhe, yaba nesandla ekuthuthukisweni kwe-API okungenani engu-3. amasayithi, alandwe ngokuzenzakalelayo imininingwane emikhankasweni ecishe ibe ngu-15 kanye nokubekwa okungaphezu kuka-000, aqoqe izimpendulo eziningi kubasebenzisi ngokusebenza komkhiqizo futhi akwazi ukushintsha inqubo eyinhloko yomkhiqizo izikhathi ezimbalwa, ngokusekelwe kule mpendulo.
I-Solution Architecture 2.0
Sekudlule iminyaka emibili kwaqala intuthuko D1.Idijithali. Ukwanda okuqhubekayo komthwalo ohlelweni kanye nokuvela kwemithombo yedatha eyengeziwe kancane kancane kwembula izinkinga ekwakhiweni kwezixazululo ezikhona.
Inkinga yokuqala ihlobene nenani ledatha elandwe kumasayithi. Sasibhekene neqiniso lokuthi ukuqoqa nokuvuselela yonke idatha edingekayo kumasayithi amakhulu kwaqala ukuthatha isikhathi esiningi kakhulu. Isibonelo, ukuqoqa idatha kusuka kusistimu yokukhangisa ye-AdRiver, esilandelela ngayo izibalo zokubekwa okuningi, kuthatha cishe amahora angu-12.
Ukuze sixazulule le nkinga, siqale ukusebenzisa zonke izinhlobo zemibiko ukuze silande idatha kumasayithi, sizama ukuthuthukisa i-API yawo kanye namasayithi ukuze isivinini sokusebenza kwayo sihlangabezane nezidingo zethu, futhi sihambisane nokulandwa kwedatha ngangokunokwenzeka.
Enye inkinga ihlobene nokucutshungulwa kwedatha elandiwe. Manje, lapho izibalo zokubeka ezintsha zifika, kuqalwa inqubo yezigaba eziningi yokubala kabusha amamethrikhi, okuhlanganisa ukulayisha idatha eluhlaza, ukubala amamethrikhi ahlanganisiwe esayithi ngalinye, ukuqhathanisa idatha evela emithonjeni ehlukene nenye, nokubala amamethrikhi esifinyezo omkhankaso. Lokhu kubangela umthwalo omningi ohlelweni lwewebhu elenza zonke izibalo. Izikhathi eziningana, phakathi nenqubo yokubala kabusha, uhlelo lokusebenza ludle yonke inkumbulo kuseva, cishe i-10-15 GB, eyayinomthelela omubi kakhulu emsebenzini wabasebenzisi ngohlelo.
Izinkinga ezihlonziwe kanye nezinhlelo zokuvelela zokuthuthukiswa okuqhubekayo komkhiqizo zisiholele esidingweni sokubheka kabusha isakhiwo sohlelo lokusebenza.
Siqale ngezixhumi.
Siqaphele ukuthi zonke izixhumi zisebenza ngokuvumelana nemodeli efanayo, ngakho-ke sakhe uhlaka lwamapayipi lapho udala khona isixhumi bekufanele uhlele kuphela ukulandelana kwezinyathelo, okunye kube okwasemhlabeni wonke. Uma isixhumi esithile sidinga ukuthuthukiswa, khona-ke sisidlulisela ngokushesha kuhlaka olusha ngesikhathi esifanayo njengoba isixhumi sithuthukiswa.
Ngesikhathi esifanayo, saqala ukuthumela izixhumi ku-Docker ne-Kubernetes.
Sihlele ukuthuthela ku-Kubernetes isikhathi eside impela, sazama izilungiselelo ze-CI/CD, kodwa saqala ukunyakaza kuphela lapho isixhumi esisodwa, ngenxa yephutha, siqala ukudla inkumbulo engaphezu kuka-20 GB kuseva, sabulala ezinye izinqubo. . Ngesikhathi sophenyo, isixhumi sithuthelwe kwiqoqo le-Kubernetes, lapho lagcina lihlala khona, ngisho nangemva kokuba iphutha selilungisiwe.
Ngokushesha saqaphela ukuthi i-Kubernetes yayifaneleka, futhi phakathi nezinyanga eziyisithupha sadlulisela izixhumi ezingu-7 kanye Nommeleli Wezixhumanisi, okudla izinsiza eziningi, kuqoqo lokukhiqiza.
Ngokulandela izixhumi, sinqume ukushintsha ukwakheka kwalo lonke uhlelo lokusebenza.
Inkinga enkulu bekungukuthi idatha isuka kuzixhumi iye kuma-proxies ngamaqoqo amakhulu, bese ishaya i-DANBoID bese ithunyelwa kuhlelo lokusebenza lwewebhu olumaphakathi ukuze lucutshungulwe. Ngenxa yenani elikhulu lokubala kabusha kwamamethrikhi, kunomthwalo omkhulu ohlelweni lokusebenza.
Kuphinde kwabonakala kunzima kakhulu ukuqapha isimo semisebenzi yokuqoqwa kwedatha ngayinye kanye nokubika amaphutha enzeke phakathi kwezixhumi ohlelweni lokusebenza lwewebhu olumaphakathi ukuze abasebenzisi babone ukuthi kwenzekani nokuthi kungani idatha ingaqoqwa.
Ukuxazulula lezi zinkinga, sithuthukise i-architecture 2.0.
Umehluko omkhulu phakathi kwenguqulo entsha yezakhiwo ukuthi esikhundleni se-Web API, sisebenzisa i-RabbitMQ kanye nelabhulali ye-MassTransit ukushintshanisa imilayezo phakathi kwezinsizakalo. Ukuze senze lokhu, kudingeke ukuthi sicishe sibhale kabusha uMmeleli Wezixhumi ngokuphelele, siwenze Ihabhu Lezixhumi. Igama lishintshiwe ngenxa yokuthi indima eyinhloko yesevisi ayisekho ekudluliseleni izicelo kuzixhumi noma emuva, kodwa ekulawuleni ukuqoqwa kwamamethrikhi asuka kuzixhumi.
Kuhlelo lokusebenza lwewebhu olumaphakathi, sihlukanise imininingwane mayelana nokubekwa nezibalo kumasayithi sazenza izinsiza ezihlukene, okwenze kwaba nokwenzeka ukususa ukubala kabusha okungadingekile futhi sigcine kuphela izibalo esezibaliwe nezihlanganisiwe ezingeni lokubeka. Siphinde sabhala kabusha futhi sathuthukisa ingqondo yokubala izibalo eziyisisekelo ngokusekelwe kudatha eluhlaza.
Ngesikhathi esifanayo, sithuthela zonke izinsiza nezinhlelo zokusebenza ku-Docker ne-Kubernetes ukuze senze isisombululo sibe lula ukusikala futhi kube lula ukusiphatha.
Sikuphi manje
I-Proof-of-concept architecture 2.0 umkhiqizo D1.Idijithali ilungile futhi isebenza endaweni yokuhlola enesethi elinganiselwe yezixhumi. Osekusele ukuthi ukwenze ukubhala kabusha ezinye izixhumi ezingu-20 kunkundla entsha, ukuhlola ukuthi idatha ilayishwe ngendlela efanele futhi wonke amamethrikhi abalwa ngendlela efanele, bese ukhiphela wonke umklamo ekukhiqizeni.
Eqinisweni, le nqubo izokwenzeka kancane kancane futhi kuzodingeka sishiye ukuhambisana emuva nama-API amadala ukuze sigcine yonke into isebenza.
Izinhlelo zethu ezisheshayo zihlanganisa ukuthuthukiswa kwezixhumi ezintsha, ukuhlanganiswa namasistimu amasha kanye nokwengeza amamethrikhi engeziwe kusethi yedatha elandwe kumasayithi axhunyiwe namasistimu okukhangisa.
Futhi sihlela ukudlulisa zonke izinhlelo zokusebenza, okuhlanganisa nesicelo sewebhu esimaphakathi, ku-Docker naku-Kubernetes. Kuhlanganiswe nesakhiwo esisha, lokhu kuzokwenza kube lula kakhulu ukuthunyelwa, ukuqapha nokulawula izinsiza ezisetshenzisiwe.
Omunye umqondo uwukuhlola ukukhetha kwesizindalwazi sokugcina izibalo, okwamanje esigcinwe ku-MongoDB. Sesivele sidlulisele izixhumi ezimbalwa ezintsha kusizindalwazi se-SQL, kodwa lapho umehluko cishe awubonakali, futhi ngezibalo ezihlanganisiwe ngosuku, ezingacelwa isikhathi esinqunyiwe, inzuzo ingaba nkulu kakhulu.
Ngokuvamile, izinhlelo zinhle kakhulu, ake siqhubeke :)
Ababhali bendatshana ethi R&D Dentsu Aegis Network Russia: Georgy Ostapenko (), Mikhail Kotsik ()
Source: www.habr.com