Inkokhelo enkulu yedatha enkulu: mayelana ne-BigData ku-telecom

Ngo-2008, i-BigData kwakuyitemu elisha nemfashini. Ngo-2019, i-BigData iyinto ethengiswayo, umthombo wenzuzo kanye nesizathu sezikweletu ezintsha.

Ngekwindla edlule, uhulumeni waseRussia uqale umthethosivivinywa wokulawula idatha enkulu. Umuntu ngamunye angeke akhonjwe ngolwazi, kodwa angakwenza lokho ngesicelo seziphathimandla zikahulumeni. Ukucubungula i-BigData yezinkampani zangaphandle kuphela ngemva kokwaziswa kwe-Roskomnadzor. Izinkampani ezinamakheli enethiwekhi angaphezu kwezinkulungwane eziyi-100 ziwela ngaphansi komthetho. Futhi, yiqiniso, lapho ngaphandle kwamarejista - kufanele idale eyodwa enohlu lwabasebenzisi be-database. Futhi uma ngaphambi kokuba le Data Enkulu ingathathwanga ngokungathi sína yiwo wonke umuntu, manje kuzofanele kucatshangelwe.

Mina, njengomqondisi wenkampani kanjiniyela okhokhayo ecubungula le datha Enkulu kakhulu, angikwazi ukuziba isizindalwazi. Ngizocabanga ngedatha enkulu ngokusebenzisa i-prism yama-opharetha e-telecom, okudlula izinhlelo zabo zokukhokha zolwazi mayelana nezinkulungwane zababhalisile nsuku zonke.

I-Theorem

Ake siqale, njengasenkingeni yezibalo: okokuqala sifakazela ukuthi idatha yabasebenzisi be-telecom ingabizwa ngokuthi i-BigDat. Ngokuvamile, idatha enkulu ibonakala ngezici ezintathu ze-VVV, nakuba ekuchazeni kwamahhala inani elithi "Vs" lifinyelele eziyisikhombisa.

Ivolumu. I-MVNO ye-Rostelecom iyodwa isebenzela ababhalisile abangaphezu kwesigidi. Abasingathi ababalulekile baphatha idatha yabantu abayizigidi ezingama-44 kuye kwezingama-78. I-traffic ikhula njalo ngomzuzwana: engxenyeni yokuqala ka-2019, ababhalisile sebevele bafinyelele u-3,3 billion GB omakhalekhukhwini.

Isivinini. Akekho ongakutshela mayelana namandla angcono kunezibalo, ngakho-ke ngizodlula kwizibikezelo ze-Cisco. Ngo-2021, u-20% wethrafikhi ye-IP uzoya kuthrafikhi yeselula - izocishe iphindwe kathathu eminyakeni emihlanu. Ingxenye yesithathu yokuxhumana kweselula izoba yi-M2M - ukuthuthukiswa kwe-IoT kuzoholela ekwandeni kokuxhumeka okuphindwe kasithupha. I-inthanethi Yezinto ngeke ibe nenzuzo kuphela, kodwa futhi izosebenzisa kakhulu izinsiza, ngakho-ke abanye opharetha bazogxila kuyo kuphela. Futhi labo abathuthukisa i-IoT njengenkonzo ehlukile bazothola ithrafikhi ephindwe kabili.

Izinhlobonhlobo. Ukuhlukahluka kuwumqondo ozimele, kodwa opharetha bezokuxhumana bazi ngempela cishe yonke into mayelana nababhalisile babo. Kusukela kugama nemininingwane yephasipoti kuya kumodeli yefoni, okuthengiwe, izindawo ezivakashelwe nezithakazelo. Ngokomthetho we-Yarovaya, amafayela emidiya agcinwa izinyanga eziyisithupha. Ngakho-ke ake sikuthathe njenge-axiom yokuthi idatha eqoqiwe iyahlukahluka.

Isoftware kanye nendlela yokusebenza

Abahlinzeki bangabanye babathengi abakhulu be-BigData, ngakho-ke izindlela ezinkulu zokuhlaziya idatha ziyasebenza embonini yezokuxhumana. Omunye umbuzo ukuthi ubani olungele ukutshala imali ekuthuthukisweni kwe-ML, i-AI, i-Deep Learning, ukutshala izimali ezikhungweni zedatha kanye nokumbiwa kwedatha. Umsebenzi ogcwele ngokugcwele onesizindalwazi uqukethe ingqalasizinda kanye nethimba, izindleko okungeyena wonke umuntu ongakwazi ukuzikhokhela. Amabhizinisi asevele anendawo yokugcina impahla yebhizinisi noma asakha indlela yokuphatha idatha kufanele abheje ku-BigData. Kulabo abangakalungeli ukutshalwa kwezimali isikhathi eside, ngikweluleka ukuthi kancane kancane wakhe i-architecture yesofthiwe futhi ufake izingxenye ngayinye ngayinye. Ungashiya amamojula asindayo kanye ne-Hadoop okokugcina. Bambalwa abantu abathenga isixazululo esenziwe ngomumo sezinkinga ezifana Nekhwalithi Yedatha kanye Nokumbiwa Kwedatha; izinkampani ngokuvamile zenza uhlelo ngendlela oyifisayo luhambisane nezicaciso nezidingo zazo ezithile - ngokwazo noma ngosizo lonjiniyela.

Kodwa akuzona zonke izinkokhelo ezingashintshwa ukuze zisebenze ne-BigData. Noma kunalokho, akuyona yonke into engashintshwa kuphela. Bambalwa abantu abangakwenza lokhu.

Izimpawu ezintathu zokuthi isistimu yokukhokha inethuba lokuba ithuluzi lokucubungula isizindalwazi:

• I-scalability evundlile. Isoftware kumele ivumelane nezimo - sikhuluma ngedatha enkulu. Ukwenyuka kwenani lolwazi kufanele kuphathwe ngokukhula ngokulinganayo kwehadiwe kuqoqo.
• Ukubekezelela amaphutha. Amasistimu akhokha kusengaphambilini anzima ngokuvamile ayakwazi ukubekezelela amaphutha ngokuzenzakalela: ukukhokhisa kufakwa kuqoqo kuma-geolocations ambalwa ukuze aqinisekisane ngokuzenzakalelayo. Kufanele futhi kube namakhompyutha anele kuqoqo le-Hadoop uma kwenzeka eyodwa noma ngaphezulu ehluleka.
• Indawo. Idatha kufanele igcinwe futhi icutshungulwe kuseva eyodwa, ngaphandle kwalokho ungaphuka ekudlulisweni kwedatha. Enye yezinhlelo ezithandwayo ze-Map-Reduce approach: Izitolo ze-HDFS, izinqubo ze-Spark. Ngokufanelekile, isofthiwe kufanele ihlanganiswe ngaphandle komthungo nengqalasizinda yesikhungo sedatha futhi ikwazi ukwenza izinto ezintathu kokukodwa: ukuqoqa, ukuhlela nokuhlaziya ulwazi.

Ithimba

Yini, kanjani futhi ngayiphi injongo uhlelo oluzocubungula idatha enkulu inqunywa yiqembu. Ngokuvamile kuqukethe umuntu oyedwa - usosayensi wedatha. Nakuba, ngokubona kwami, inani elincane labasebenzi le-Big Data lihlanganisa nomphathi womkhiqizo, unjiniyela wedatha, kanye nomphathi. Owokuqala uqonda izinsizakalo, uhumushela ulimi lobuchwepheshe olimini lwabantu futhi ngokuphambene nalokho. Unjiniyela Wedatha uphilisa amamodeli esebenzisa i-Java/Scala kanye nokuhlola ngokuFunda ngomshini. Umphathi uqondisa, abeke imigomo, futhi alawule izigaba.

Izinkinga

Kusohlangothini lwethimba le-BigData lapho izinkinga zivame ukuvela lapho kuqoqwa futhi kucutshungulwa idatha. Uhlelo ludinga ukuchaza ukuthi yini okufanele iqoqwe nokuthi icutshungulwa kanjani - ukuze uchaze lokhu, udinga ukukuqonda kuqala ngokwakho. Kodwa kubahlinzeki, izinto azilula kangako. Ngikhuluma ngezinkinga zisebenzisa isibonelo somsebenzi wokunciphisa i-churn yababhalisi - yilokhu opharetha be-telecom abazama ukuxazulula ngosizo lwe-Big Data kwasekuqaleni.

Ukubeka imigomo. Ukucaciswa kwezobuchwepheshe okubhalwe kahle kanye nokuqonda okuhlukile kwamagama kube ubuhlungu bamakhulu eminyaka hhayi kuma-freelancers kuphela. Ngisho nababhalisile "abawehlisiwe" bangahunyushwa ngezindlela ezahlukene - njengalabo abangasebenzisanga izinsizakalo zomqhubi inyanga, izinyanga eziyisithupha noma unyaka. Futhi ukuze udale i-MVP ngokusekelwe kudatha yomlando, udinga ukuqonda imvamisa yembuyiselo yababhalisi abavela ku-churn - labo abazame abanye opharetha noma bashiye idolobha futhi basebenzise inombolo ehlukile. Omunye umbuzo obalulekile: isikhathi esingakanani ngaphambi kokuthi obhalisile alindelwe ukuthi ahambe kufanele umhlinzeki anqume lokhu futhi athathe isinyathelo? Izinyanga eziyisithupha kusesekuseni kakhulu, isonto sekwephuze kakhulu.

Ukushintshwa kwemiqondo. Ngokuvamile, opharetha bakhomba iklayenti ngenombolo yocingo, ngakho-ke kunengqondo ukuthi izimpawu kufanele zilayishwe bezisebenzisa. Kuthiwani nge-akhawunti yakho yomuntu siqu noma inombolo yesicelo sesevisi? Kuyadingeka ukunquma ukuthi iyiphi iyunithi okufanele ithathwe njengeklayenti ukuze idatha ohlelweni lomqhubi ingahluka. Ukuhlola inani leklayenti nakho kuyangabazeka - yimuphi obhalisile obaluleke kakhulu enkampanini, yimuphi umsebenzisi odinga umzamo owengeziwe ukuwugcina, nokuthi yimaphi "azowa" kunoma yikuphi futhi asikho isidingo sokusebenzisa izinsiza kubo.

Ukuntuleka kolwazi. Akubona bonke abasebenzi abahlinzeki abakwazi ukuchazela ithimba le-BigData ukuthi yini ethinta ngokuqondile i-churn yababhalisile nokuthi izici ezingaba khona ekukhokheni zibalwa kanjani. Ngisho noma beqambe enye yazo - i-ARPU - kuvela ukuthi ingabalwa ngezindlela ezihlukene: ngezinkokhelo zekhasimende ngezikhathi ezithile, noma ngezindleko zokukhokha ezizenzakalelayo. Futhi ngesikhathi somsebenzi, kuphakama eminye imibuzo eyisigidi. Ingabe imodeli ihlanganisa wonke amaklayenti, iyini intengo yokugcina iklayenti, ingabe likhona iphuzu lokucabanga ngamanye amamodeli, nokuthi yini okufanele yenziwe ngamaklayenti agcinwe ngephutha.

Ukulungiselelwa umgomo. Ngazi ngezinhlobo ezintathu zamaphutha emiphumela abangela opharetha ukuthi bakhungatheke ngesizindalwazi.

1. Umhlinzeki utshala ku-BigData, ucubungula ama-gigabytes olwazi, kodwa uthola umphumela obungatholakala ngenani eliphansi. Imidwebo elula namamodeli, ama-analytics akudala asetshenziswa. Izindleko ziphakeme izikhathi eziningi, kodwa umphumela uyafana.
2. Umsebenzisi uthola idatha enezinhlangothi eziningi njengokuphumayo, kodwa akaqondi ukuthi isetshenziswa kanjani. Kukhona ama-analytics - nansi, iyaqondakala futhi i-voluminous, kodwa ayisizi ngalutho. Umphumela wokugcina, ongakwazi ukufaka umgomo "wokucubungula idatha," awukacatshangelwa. Akwanele ukucubungula - izibalo kufanele zibe yisisekelo sokubuyekeza izinqubo zebhizinisi.
3. Izithiyo ekusetshenzisweni kokuhlaziya kwe-BigData zingaba izinqubo zebhizinisi eziphelelwe yisikhathi nesofthiwe engafanele izinjongo ezintsha. Lokhu kusho ukuthi benze iphutha esigabeni sokulungiselela - abazange bacabange nge-algorithm yezenzo kanye nezigaba zokwethulwa kwe-Big Data emsebenzini.

Kungani

Ekhuluma ngemiphumela. Ngizobheka izindlela zokusebenzisa nokwenza imali ngeDatha Enkulu abaqhubi bezingcingo asebevele bezisebenzisa.
Abahlinzeki ababikezeli kuphela ukuphuma kwababhalisile, kodwa futhi nomthwalo eziteshini eziyisisekelo.

1. Ulwazi mayelana nokunyakaza kwababhalisile, umsebenzi kanye nezinsizakalo zemvamisa kuyahlaziywa. Umphumela: ukuncishiswa kwenani lokulayishwa ngokweqile ngenxa yokwenziwa ngcono kanye nokwenziwa kwesimanje kwezinkinga zengqalasizinda.
2. Opharetha be-Telecom basebenzisa ulwazi mayelana nendawo yababhalisile kanye nokuminyana kwethrafikhi lapho bevula izindawo zokuthengisa. Ngakho, izibalo ze-BigData sezivele zisetshenziswa yi-MTS kanye ne-VimpelCom ukuhlela indawo yamahhovisi amasha.
3. Abahlinzeki benza imali ngedatha yabo enkulu ngokuyinikeza izinkampani zangaphandle. Amakhasimende amakhulu ama-opharetha e-BigData amabhange okuhweba. Besebenzisa i-database, baqapha imisebenzi esolisayo ye-SIM khadi yobhalisile lapho amakhadi axhunywe khona, futhi basebenzise amaphuzu obungozi, izinsizakalo zokuqinisekisa kanye nokuqapha. Futhi ngo-2017, uhulumeni waseMoscow wacela ukunyakaza okusekelwe kudatha ye-BigData evela ku-Tele2 ukuhlela ingqalasizinda yezobuchwepheshe nezokuthutha.
4. I-BigData analytics iyimayini yegolide yabakhangisi, abangakha imikhankaso yokukhangisa yomuntu siqu yezinkulungwane zamaqembu ababhalisile uma bekhetha. Izinkampani zeTelecom zihlanganisa amaphrofayili omphakathi, izintshisekelo zabathengi namaphethini okuziphatha kwababhalisile, bese zisebenzisa i-BigData eqoqiwe ukuheha amakhasimende amasha. Kodwa ngokuphromotha okukhulu kanye nokuhlelwa kwe-PR, ukukhokhisa akuhlali kunokusebenza okwanele: uhlelo kufanele ngesikhathi esisodwa lucabangele izici eziningi ngokuhambisana nolwazi oluningiliziwe mayelana namaklayenti.

Ngenkathi abanye besabheka i-BigData njengebinzana elingenalutho, i-Big Four vele yenza imali ngayo. I-MTS ithola ama-ruble ayizigidi eziyizinkulungwane ezingu-14 ngokucutshungulwa kwedatha enkulu ezinyangeni eziyisithupha, futhi i-Tele2 ikhuphule imali engenayo evela kumaphrojekthi izikhathi ezintathu nengxenye. I-BigData iyashintsha isuka kuthrendi ibe into okufanele ube nayo, lapho kuzokwakhiwa khona kabusha lonke uhlaka lwabasebenza ngocingo.

Source: www.habr.com