{"id":6739,"date":"2022-07-01T13:43:11","date_gmt":"2022-07-01T10:43:11","guid":{"rendered":"https:\/\/cloudfresh.com\/?p=6739"},"modified":"2025-11-27T18:25:20","modified_gmt":"2025-11-27T15:25:20","slug":"the-future-of-data-choosing-a-data-platform-and-best-practices-of-data-management","status":"publish","type":"post","link":"https:\/\/cloudfresh.com\/cs\/blog\/budoucnost-dat-vyber-datove-platformy-a-osvedcene-postupy-spravy-dat\/","title":{"rendered":"Budoucnost dat: V\u00fdb\u011br datov\u00e9 platformy a osv\u011bd\u010den\u00e9 postupy spr\u00e1vy dat"},"content":{"rendered":"

\"\"<\/p>\n

 <\/p>\n

Podle ned\u00e1vn\u00e9ho v\u00fdzkumu datov\u00e9 platformy Google Cloud<\/a> budou data v budoucnu jednotn\u00e1, flexibiln\u00ed a snadno dostupn\u00e1.<\/p>\n

Data maj\u00ed z\u00e1sadn\u00ed v\u00fdznam pro \u0159\u00edzen\u00ed inovativn\u00edch rozhodnut\u00ed o produktech a u\u017eivatelsk\u00fdch zku\u0161enostech spolu s rozs\u00e1hl\u00fdmi strategiemi uv\u00e1d\u011bn\u00ed na trh. \u00dasp\u011b\u0161n\u00e9 vyu\u017eit\u00ed dat v\u00e1m m\u016f\u017ee poskytnout v\u00fdznamnou konkuren\u010dn\u00ed v\u00fdhodu. Proto v\u011bt\u0161ina technologick\u00fdch spole\u010dnost\u00ed a startup\u016f investuje do spr\u00e1vy dat – aby se modernizovaly a fungovaly ve st\u00e1le v\u011bt\u0161\u00edm m\u011b\u0159\u00edtku, aby ospravedlnily sou\u010dasn\u00e9 i budouc\u00ed n\u00e1klady na data a aby pozvedly svou organiza\u010dn\u00ed vysp\u011blost a rozhodov\u00e1n\u00ed.<\/p>\n

 <\/p>\n

Podle pr\u016fzkumu Google Cloud existuj\u00ed 3 kl\u00ed\u010dov\u00e9 p\u0159\u00edstupy k dat\u016fm, kter\u00fdch se inovativn\u00ed technologick\u00e9 spole\u010dnosti dr\u017e\u00ed:<\/h4>\n

 <\/p>\n

    \n
  1. Data mus\u00ed b\u00fdt sjednocena nap\u0159\u00ed\u010d celou spole\u010dnost\u00ed, a to dokonce i mezi dodavateli a partnery.<\/li>\n
  2. Technologick\u00fd stack mus\u00ed b\u00fdt dostate\u010dn\u011b flexibiln\u00ed, aby podporoval r\u016fzn\u00e9 p\u0159\u00edpady pou\u017eit\u00ed od offline anal\u00fdzy dat a\u017e po strojov\u00e9 u\u010den\u00ed v re\u00e1ln\u00e9m \u010dase.<\/li>\n
  3. Stack mus\u00ed b\u00fdt tak\u00e9 snadno dostupn\u00fd a mus\u00ed podporovat r\u016fzn\u00e9 platformy, programovac\u00ed jazyky, n\u00e1stroje a otev\u0159en\u00e9 standardy.<\/li>\n<\/ol>\n

     <\/p>\n

    S p\u0159\u00edstupem, ukl\u00e1d\u00e1n\u00edm, nekonzistentn\u00edmi n\u00e1stroji, dodr\u017eov\u00e1n\u00edm p\u0159edpis\u016f a zabezpe\u010den\u00edm jsou v\u0161ak spojeny probl\u00e9my, kter\u00e9 zt\u011b\u017euj\u00ed proniknut\u00ed pod povrch a uvoln\u011bn\u00ed skute\u010dn\u00e9 hodnoty z va\u0161ich dat. Mezi n\u011b pat\u0159\u00ed nap\u0159:<\/h4>\n

     <\/p>\n

      \n
    1. Zd\u011bd\u011bn\u00e9 star\u0161\u00ed ekosyst\u00e9my s r\u016fzn\u00fdmi technologick\u00fdmi soubory<\/li>\n
    2. Rozhodnut\u00ed, zda ukl\u00e1dat data v jednom nebo v\u00edce cloudech.<\/li>\n
    3. D\u00e1vkov\u00e1n\u00ed nebo mikrod\u00e1vkov\u00e1n\u00ed va\u0161ich dat dnes nam\u00edsto jejich zpracov\u00e1n\u00ed v re\u00e1ln\u00e9m \u010dase<\/li>\n
    4. Chyb\u011bj\u00edc\u00ed snadn\u00fd p\u0159\u00edstup ke v\u0161em va\u0161im dat\u016fm a chyb\u011bj\u00edc\u00ed mo\u017enost jejich zpracov\u00e1n\u00ed a anal\u00fdzy<\/li>\n<\/ol>\n

      <\/h4>\n

      Doporu\u010dujeme dv\u011b hlavn\u00ed z\u00e1sady pro v\u00fdb\u011br datov\u00e9 platformy, kter\u00e1 v\u00e1m pom\u016f\u017ee vy\u0159e\u0161it probl\u00e9my s daty a posunout spr\u00e1vu dat na vy\u0161\u0161\u00ed \u00farove\u0148.<\/h4>\n

       <\/p>\n

      Z\u00e1sada 1: Jednoduchost a \u0161k\u00e1lovatelnost (m\u011b\u0159itelnost)<\/h2>\n

       <\/p>\n

      Men\u0161\u00ed syst\u00e9my jsou obecn\u011b jednodu\u0161\u0161\u00ed. U\u017e v\u0161ak nemus\u00edte volit mezi syst\u00e9mem, kter\u00fd se snadno pou\u017e\u00edv\u00e1, a syst\u00e9mem, kter\u00fd je vysoce \u0161k\u00e1lovateln\u00fd. Pou\u017eit\u00ed architektury bez server\u016f eliminuje pot\u0159ebu spr\u00e1vy cluster\u016f a d\u00e1v\u00e1 v\u00e1m mo\u017enost zvl\u00e1dnout masivn\u00ed \u0161k\u00e1lov\u00e1n\u00ed jak pro v\u00fdpo\u010detn\u00ed, tak pro \u00falo\u017en\u00e9 syst\u00e9my, tak\u017ee se u\u017e nikdy nebudete muset ob\u00e1vat, \u017ee velikost dat p\u0159ekro\u010d\u00ed va\u0161e technick\u00e9 mo\u017enosti. Pro jednoduchost i \u0161k\u00e1lovatelnost doporu\u010dujeme datovou platformu bez serveru. Tak\u00e9 doporu\u010dujeme vy\u0159adit v\u0161echny mo\u017enosti, kter\u00e9 vy\u017eaduj\u00ed instalaci softwaru, spr\u00e1vu cluster\u016f nebo lad\u011bn\u00ed dotaz\u016f.<\/p>\n

       <\/p>\n

      Z\u00e1sada 2: Agilita a sni\u017eov\u00e1n\u00ed n\u00e1klad\u016f<\/h2>\n

       <\/p>\n

      Jak\u00fdkoli syst\u00e9m spr\u00e1vy dat, kter\u00fd kombinuje v\u00fdpo\u010detn\u00ed a \u00falo\u017en\u00fd syst\u00e9m, v\u00e1s bude nutit zvy\u0161ovat v\u00fdpo\u010detn\u00ed v\u00fdkon, abyste se vypo\u0159\u00e1dali s rostouc\u00edm objemem dat, i kdy\u017e to nebudete pot\u0159ebovat. To m\u016f\u017ee b\u00fdt n\u00e1kladn\u00e9 a m\u016f\u017ee se st\u00e1t, \u017ee budete d\u011blat kompromisy, nap\u0159\u00edklad ukl\u00e1dat do analytick\u00e9ho skladu pouze data za posledn\u00edch dvan\u00e1ct m\u011bs\u00edc\u016f.<\/p>\n

      Chcete-li co nejv\u00edce omezit spr\u00e1vu infrastruktury, zva\u017ete bezserverov\u00fd, multicloudov\u00fd datov\u00fd sklad se zv\u00fd\u0161enou spolehlivost\u00ed, v\u00fdkonem a integrovanou ochranou dat (nap\u0159\u00edklad BigQuery<\/a>).<\/p>\n

      S n\u011b\u010d\u00edm, jako je BigQuery, nemus\u00edte p\u0159edem pl\u00e1novat dotazy ani indexovat datov\u00e9 sady. Odd\u011blen\u00e9 \u00falo\u017ei\u0161t\u011b a v\u00fdpo\u010detn\u00ed v\u00fdkon v\u00e1m umo\u017en\u00ed p\u0159istavovat data bez obav, \u017ee v\u00e1m to zv\u00fd\u0161\u00ed n\u00e1klady na dotazov\u00e1n\u00ed, a va\u0161i datov\u00ed v\u011bdci mohou experimentovat, ani\u017e by se museli starat o clustery nebo dimenzov\u00e1n\u00ed datov\u00fdch sklad\u016f, aby mohli zkou\u0161et nov\u00e9 n\u00e1pady prost\u0159ednictv\u00edm ad hoc dotaz\u016f.<\/p>\n

       <\/p>\n

      Nyn\u00ed, pot\u00e9 co jsme si pro\u0161li z\u00e1sady v\u00fdb\u011bru spr\u00e1vn\u00e9 platformy pro spr\u00e1vu dat, zd\u016frazn\u00edme n\u011bkter\u00e9 osv\u011bd\u010den\u00e9 postupy spr\u00e1vy dat:<\/h4>\n

       <\/p>\n

      Rozhodov\u00e1n\u00ed na z\u00e1klad\u011b dat v re\u00e1ln\u00e9m \u010dase<\/h2>\n

       <\/p>\n

      Chcete b\u00fdt schopni zachytit data v re\u00e1ln\u00e9m \u010dase a zp\u0159\u00edstupnit je obchodn\u00edm t\u00fdm\u016fm pro dotazov\u00e1n\u00ed s n\u00edzkou latenc\u00ed? Chcete se tak\u00e9 ujistit, \u017ee va\u0161e streamovac\u00ed kan\u00e1ly jsou \u0161k\u00e1lovateln\u00e9, odoln\u00e9 a maj\u00ed n\u00edzkou re\u017eii spr\u00e1vy? BigQuery m\u00e1 nativn\u00ed podporu pro p\u0159ij\u00edm\u00e1n\u00ed streamovan\u00fdch dat a okam\u017eit\u011b je zp\u0159\u00edstup\u0148uje pro anal\u00fdzu pomoc\u00ed jazyka SQL. Spolu se snadno pou\u017eiteln\u00fdm rozhran\u00edm API pro streamov\u00e1n\u00ed BigQuery v\u00e1m Dataflow d\u00e1v\u00e1 mo\u017enost spravovat sez\u00f3nn\u00ed a prudce rostouc\u00ed pracovn\u00ed z\u00e1t\u011b\u017e bez nadm\u011brn\u00fdch n\u00e1klad\u016f.<\/p>\n

       <\/p>\n

      Odbour\u00e1v\u00e1n\u00ed datov\u00fdch sil<\/h2>\n

       <\/p>\n

      V mnoha organizac\u00edch nakonec doch\u00e1z\u00ed k vytv\u00e1\u0159en\u00ed datov\u00fdch sil, proto\u017ee ukl\u00e1daj\u00ed data odd\u011blen\u011b nap\u0159\u00ed\u010d odd\u011blen\u00edmi a obchodn\u00edmi jednotkami, p\u0159i\u010dem\u017e ka\u017ed\u00fd t\u00fdm vlastn\u00ed sv\u00e1 vlastn\u00ed data. To znamen\u00e1, \u017ee kdykoli chcete prov\u00e9st anal\u00fdzu, mus\u00edte p\u0159ij\u00edt na to, jak tato sila rozb\u00edt. Dne\u0161n\u00ed multicloudov\u00e1 a hybridn\u011b-cloudov\u00e1 realita vy\u017eaduje dal\u0161\u00ed \u00farove\u0148 sofistikovanosti p\u0159i spr\u00e1v\u011b a p\u0159\u00edstupu k odd\u011blen\u00fdm dat\u016fm.<\/p>\n

      V\u0161echna data m\u016f\u017eete um\u00edstit do BigQuery a poskytnout opakovan\u011b pou\u017eiteln\u00e9 funkce, materializovan\u00e9 pohledy, a dokonce i mo\u017enost tr\u00e9novat ML modely bez jak\u00e9hokoli p\u0159esunu dat. To znamen\u00e1, \u017ee i netechni\u010dt\u00ed experti na danou oblast (a partne\u0159i a dodavatel\u00e9, kte\u0159\u00ed k tomu maj\u00ed opr\u00e1vn\u011bn\u00ed) mohou snadno p\u0159istupovat k dat\u016fm a pou\u017e\u00edvat SQL k dotazov\u00e1n\u00ed pomoc\u00ed zn\u00e1m\u00fdch n\u00e1stroj\u016f, jako jsou tabulky a ovl\u00e1dac\u00ed panely.<\/p>\n

       <\/p>\n

      Usnadn\u011bn\u00ed p\u0159\u00edstupu ke v\u0161em dat\u016fm<\/h2>\n

       <\/p>\n

      Historicky se nestrukturovan\u00e1 a \u010d\u00e1ste\u010dn\u011b strukturovan\u00e1 data nejl\u00e9pe hodila do datov\u00fdch jezer, zat\u00edmco strukturovan\u00e1 data do datov\u00fdch sklad\u016f. Toto rozd\u011blen\u00ed vytv\u00e1\u0159elo technologick\u00e1 sila, kter\u00e1 zt\u011b\u017eovala p\u0159ekon\u00e1n\u00ed form\u00e1tov\u00e9ho rozd\u011blen\u00ed; v\u0161echna data jste ukl\u00e1dali do datov\u00e9ho jezera, proto\u017ee je to levn\u011bj\u0161\u00ed a snadn\u011bji se spravuj\u00ed, a pak jste data p\u0159esunuli do datov\u00e9ho skladu, abyste mohli pou\u017e\u00edt analytick\u00e9 n\u00e1stroje k z\u00edsk\u00e1n\u00ed poznatk\u016f.<\/p>\n

       <\/p>\n

      <\/h2>\n

      Vyu\u017eit\u00ed AI\/ML k rychlej\u0161\u00edmu experimentov\u00e1n\u00ed a spr\u00e1v\u011b pracovn\u00edch z\u00e1t\u011b\u017e\u00ed<\/h2>\n

       <\/p>\n

      Pokud to s diferenciac\u00ed na z\u00e1klad\u011b dat mysl\u00edte v\u00e1\u017en\u011b, chcete ze shroma\u017e\u010fovan\u00fdch dat vyt\u011b\u017eit co nejvy\u0161\u0161\u00ed hodnotu. Abyste toho dos\u00e1hli, chcete, aby va\u0161e t\u00fdmy datov\u00e9 v\u011bdy byly co nejproduktivn\u011bj\u0161\u00ed a neprop\u00e1sly \u017e\u00e1dn\u00e9 p\u0159\u00edle\u017eitosti.<\/p>\n

      Kvalita va\u0161ich p\u0159edp\u0159ipraven\u00fdch a n\u00edzkok\u00f3dov\u00fdch model\u016f je kl\u00ed\u010dov\u00e1. AutoML na Vertex AI zp\u0159\u00edstup\u0148uje nejlep\u0161\u00ed modely AI ve sv\u00e9 t\u0159\u00edd\u011b v prost\u0159ed\u00ed bez k\u00f3du, co\u017e umo\u017e\u0148uje rychl\u00e9 srovn\u00e1v\u00e1n\u00ed a ur\u010dov\u00e1n\u00ed priorit.<\/p>\n

      Aby syst\u00e9my v produkci p\u0159in\u00e1\u0161ely skute\u010dnou hodnotu, mus\u00ed b\u00fdt schopny p\u0159ij\u00edmat, zpracov\u00e1vat a obsluhovat data a strojov\u00e9 u\u010den\u00ed mus\u00ed \u0159\u00eddit personalizovan\u00e9 slu\u017eby v re\u00e1ln\u00e9m \u010dase na z\u00e1klad\u011b kontextu z\u00e1kazn\u00edka.<\/p>\n

       <\/p>\n

      Hodn\u011b jsme mluvili o vyu\u017eit\u00ed dat a o tom, co to vlastn\u011b znamen\u00e1, spolu s n\u011bkter\u00fdmi \u00favahami, kter\u00fdm m\u016f\u017eete \u010delit p\u0159i migraci na datov\u00fd sklad v cloudu.<\/p>\n

      Chcete-li se dozv\u011bd\u011bt v\u00edce o tom, jak v\u00e1m Google Cloud m\u016f\u017ee pomoci vyu\u017e\u00edt poznatky k z\u00edsk\u00e1n\u00ed v\u00fdznamn\u00e9 obchodn\u00ed v\u00fdhody – m\u016f\u017eete se obr\u00e1tit na ofici\u00e1ln\u00edho partnera Google Cloud Premier Partner – Cloudfresh.<\/p>\n

       <\/p>\n

      T\u00fdm Cloudfresh je jedine\u010dn\u00fdm centrem odborn\u00fdch znalost\u00ed pro slu\u017eby Google Cloud<\/a>, Zendesk<\/a> a Asana<\/a>. Pro tyto produkty v\u00e1m m\u016f\u017eeme poskytnout n\u00e1sleduj\u00edc\u00ed slu\u017eby:<\/p>\n