Moni nonse! Dzina langa ndine Golov Nikolay. Poyamba, ndinagwira ntchito ku Avito ndikuyendetsa Data Platform kwa zaka zisanu ndi chimodzi, ndiko kuti, ndinagwira ntchito pazinthu zonse: analytical (Vertica, ClickHouse), kusindikiza ndi OLTP (Redis, Tarantool, VoltDB, MongoDB, PostgreSQL). Panthawiyi, ndidakumana ndi ma database ambiri - osiyana kwambiri komanso osazolowereka, komanso osagwiritsidwa ntchito mokhazikika.
Panopa ndikugwira ntchito ku ManyChat. M'malo mwake, ichi ndi chiyambi - chatsopano, chokhumba komanso kukula mofulumira. Ndipo nditayamba kulowa nawo kampaniyo, funso lachikale lidawuka: "Kodi woyambitsa wachinyamata atenge chiyani kuchokera ku DBMS ndi msika wa database?"
M'nkhaniyi, kutengera lipoti langa pa , ndiyankha funso ili. Kanema wa lipotilo akupezeka pa .
Ma database odziwika bwino a 2020
Ndi 2020, ndidayang'ana ndikuwona mitundu itatu yama database.
Mtundu woyamba - Zolemba zakale za OLTP: PostgreSQL, SQL Server, Oracle, MySQL. Zinalembedwa kalekale, koma ndi zofunikabe chifukwa ndizodziwika bwino kwa anthu otukula.
Mtundu wachiwiri - maziko kuchokera ku "zero". Adayesa kuchoka pamachitidwe akale posiya SQL, zomanga zachikhalidwe ndi ACID, powonjezera sharding ndi zina zowoneka bwino. Mwachitsanzo, izi ndi Cassandra, MongoDB, Redis kapena Tarantool. Mayankho onsewa amafuna kupatsa msika china chatsopano komanso chokhazikika chifukwa adakhala osavuta kuchita zina. Ndiwonetsa nkhokwe izi ndi ambulera ya NOSQL.
"Zero" zatha, tidazolowera nkhokwe za NOSQL, ndipo dziko, monga momwe ndimawonera, lidatenga gawo lotsatira - nkhokwe zoyendetsedwa. Ma databasewa ali ndi maziko ofanana ndi ma database akale a OLTP kapena atsopano a NoSQL. Koma alibe chifukwa cha DBA ndi DevOps ndipo amayendetsa pa hardware yoyendetsedwa pamitambo. Kwa wopanga mapulogalamu, izi ndi "zoyambira" zomwe zimagwira ntchito kwinakwake, koma palibe amene amasamala momwe zimayikidwira pa seva, yemwe adakonza seva ndikuyisintha.
Zitsanzo za nkhokwe zotere:
- AWS RDS ndi cholembera choyendetsedwa cha PostgreSQL/MySQL.
- DynamoDB ndi analogue ya AWS ya database yochokera ku zolemba, zofanana ndi Redis ndi MongoDB.
- Amazon Redshift ndi nkhokwe yowunikira.
Izi ndizosungira zakale, koma zokulira m'malo oyendetsedwa, osafunikira kugwira ntchito ndi zida.
Zindikirani. Zitsanzo zimatengedwa ku chilengedwe cha AWS, koma ma analogue awo amapezekanso ku Microsoft Azure, Google Cloud, kapena Yandex.Cloud.
Chatsopano ndi chiyani pa izi? Mu 2020, palibe izi.
Lingaliro lopanda seva
Chatsopano pamsika mu 2020 ndi mayankho opanda seva kapena opanda seva.
Ndiyesera kufotokoza zomwe izi zikutanthauza kugwiritsa ntchito chitsanzo cha utumiki wokhazikika kapena ntchito yobwerera.
Kuti titumize pulogalamu yanthawi zonse ya backend, timagula kapena kubwereka seva, kukopera kachidindoyo, kusindikiza pomaliza kunja ndikulipira renti, magetsi ndi ma data center. Ichi ndiye chiwembu chokhazikika.
Kodi pali njira ina? Ndi ntchito zopanda seva mungathe.
Kodi cholinga cha njirayi ndi chiyani: palibe seva, palibe ngakhale kubwereka zochitika mumtambo. Kuti mutumize ntchitoyo, lembani kachidindo (ntchito) kunkhokwe ndikusindikiza mpaka kumapeto. Kenako timangolipira kuyitana kulikonse ku ntchitoyi, kunyalanyaza kwathunthu zida zomwe zimachitidwa.
Ndiyesera kufotokoza njira iyi ndi zithunzi.
Kutumiza kwachikale. Tili ndi ntchito yokhala ndi katundu wina wake. Timakweza zochitika ziwiri: ma seva akuthupi kapena zochitika mu AWS. Zopempha zakunja zimatumizidwa kuzochitika izi ndikukonzedwa pamenepo.
Monga mukuwonera pachithunzichi, ma seva samatayidwa mofanana. Imodzi imagwiritsidwa ntchito 100%, pali zopempha ziwiri, ndipo imodzi ndi 50% yokha - yopanda ntchito. Ngati palibe zopempha zitatu zomwe zikufika, koma 30, ndiye kuti dongosolo lonse silingathe kulimbana ndi katunduyo ndipo lidzayamba kuchepa.
Kutumiza kopanda seva. M'malo opanda seva, ntchito yotereyi ilibe zochitika kapena ma seva. Pali dziwe lina lazinthu zotenthetsera - zotengera zazing'ono zokonzeka za Docker zokhala ndi ma code ogwiritsiridwa ntchito. Dongosolo limalandira zopempha zakunja ndipo kwa aliyense wa iwo chimango chopanda seva chimakweza chidebe chaching'ono chokhala ndi code: chimakonza pempholi ndikupha chidebecho.
Pempho limodzi - chidebe chimodzi chokwezeka, zopempha 1000 - zotengera 1000. Ndipo kutumizidwa pa ma seva a hardware kale ndi ntchito ya opereka mtambo. Zimabisika kwathunthu ndi dongosolo lopanda seva. Mu lingaliro ili timalipira kuyitana kulikonse. Mwachitsanzo, kuyimba kumodzi kunabwera tsiku - tidalipira kuyimba kamodzi, miliyoni idabwera pamphindi - tidalipira miliyoni. Kapena pamphindi, izi zimachitikanso.
Lingaliro la kusindikiza ntchito yopanda seva ndi yoyenera kwa ntchito yopanda malire. Ndipo ngati mukufuna (boma) ntchito yokhazikika, ndiye timawonjezera nkhokwe ku ntchitoyo. Pankhaniyi, ikafika pogwira ntchito ndi boma, ntchito iliyonse ya statefull imangolemba ndikuwerenga kuchokera ku database. Komanso, kuchokera m'nkhokwe yamtundu uliwonse mwa mitundu itatu yomwe yafotokozedwa koyambirira kwa nkhaniyi.
Kodi malire onsewa ndi otani? Izi ndi ndalama zomwe zimagwiritsidwa ntchito nthawi zonse mtambo kapena seva ya hardware (kapena ma seva angapo). Zilibe kanthu kaya timagwiritsa ntchito nkhokwe yachikale kapena yoyendetsedwa, kaya tili ndi Devops ndi admin kapena ayi, timalipirabe hardware, magetsi ndi data center 24/7. Ngati tili ndi maziko apamwamba, timalipira mbuye ndi kapolo. Ngati ndi database yodzaza kwambiri, timalipira ma seva 10, 20 kapena 30, ndipo timalipira nthawi zonse.
Kukhalapo kwa ma seva osungidwa kosatha mu mtengo wamtengo wapatali kunkawoneka ngati koyenera. Ma database ochiritsira amakhalanso ndi zovuta zina, monga malire pa kuchuluka kwa maulumikizidwe, zoletsa makulitsidwe, mgwirizano wogawidwa ndi geo - zitha kuthetsedwa mwanjira ina, koma osati zonse nthawi imodzi osati bwino.
Serverless database - chiphunzitso
Funso la 2020: ndizothekanso kupanga database yopanda seva? Aliyense wamvapo za backend yopanda seva ... tiyeni tiyese kupanga database yopanda seva?
Izi zikumveka zachilendo, chifukwa nkhokwe ndi ntchito yokhazikika, osati yoyenera kwambiri pazipangizo zopanda seva. Panthawi imodzimodziyo, malo a database ndi aakulu kwambiri: gigabytes, terabytes, ndi ma analytical databases ngakhale ma petabytes. Sikosavuta kuyikweza m'matumba opepuka a Docker.
Kumbali inayi, pafupifupi nkhokwe zamakono zonse zimakhala ndi malingaliro ambiri ndi zigawo zikuluzikulu: zochitika, kugwirizanitsa kukhulupirika, ndondomeko, kudalira maubwenzi ndi malingaliro ambiri. Pazinthu zambiri zama database, malo ochepa ndi okwanira. Ma Gigabytes ndi Terabytes amagwiritsidwa ntchito mwachindunji ndi gawo laling'ono chabe la malingaliro a database omwe akukhudzidwa mwachindunji ndi mafunso.
Chifukwa chake, lingaliro ndilakuti: ngati gawo lina lamalingaliro limalola kuphedwa kopanda malire, bwanji osagawa maziko kukhala magawo a Stateful ndi Stateless.
Zopanda seva pazayankho za OLAP
Tiyeni tiwone momwe kudula nkhokwe m'magawo a Stateful ndi Stateless kungawonekere pogwiritsa ntchito zitsanzo zothandiza.
Mwachitsanzo, tili ndi analytical database: deta yakunja (silinda yofiira kumanzere), ndondomeko ya ETL yomwe imanyamula deta mu database, ndi katswiri yemwe amatumiza mafunso a SQL ku database. Iyi ndi ndondomeko yakale yosungiramo zinthu zakale.
Muchiwembu ichi, ETL imachitika kamodzi kokha. Ndiye muyenera kulipira nthawi zonse ma seva omwe database imayenda ndi data yodzazidwa ndi ETL, kuti pakhale china chake choti mutumizeko.
Tiyeni tiwone njira ina yomwe yakhazikitsidwa mu AWS Athena Serverless. Palibe zida zodzipatulira kwamuyaya zomwe zidatsitsidwa zimasungidwa. M'malo mwa izi:
- Wogwiritsa amatumiza funso la SQL kwa Athena. Athena optimizer amasanthula funso la SQL ndikufufuza metadata store (Metadata) kuti adziwe zomwe zikufunika kuti ayankhe funsolo.
- The optimizer, kutengera zomwe zasonkhanitsidwa, imatsitsa zofunikira kuchokera kuzinthu zakunja ndikusunga kwakanthawi (zosakhalitsa).
- Funso la SQL lochokera kwa wogwiritsa ntchito limagwiritsidwa ntchito posungira kwakanthawi ndipo zotsatira zake zimabwezeredwa kwa wogwiritsa ntchito.
- Kusungirako kwakanthawi kumachotsedwa ndipo zothandizira zimatulutsidwa.
Muzomangamangazi, timangolipira njira yochitira pempholi. Palibe zopempha - palibe ndalama.
Iyi ndi njira yogwirira ntchito ndipo ikugwiritsidwa ntchito osati ku Athena Serverless, komanso ku Redshift Spectrum (mu AWS).
Chitsanzo cha Athena chikuwonetsa kuti database ya Serverless imagwira ntchito pamafunso enieni ndi makumi ndi mazana a Terabytes a data. Mazana a Terabytes adzafuna mazana a maseva, koma sitiyenera kuwalipira - timalipira zopemphazo. Kuthamanga kwa pempho lililonse ndi (kwambiri) kutsika poyerekeza ndi nkhokwe zapadera monga Vertica, koma sitimalipira nthawi yopuma.
Dongosolo loterolo limagwira ntchito pamafunso osowa ad-hoc. Mwachitsanzo, tikaganiza zongoyesa kuganiza mozama pazambiri zazikulu. Athena ndiwabwino pamilandu iyi. Pazopempha nthawi zonse, dongosolo loterolo ndi lokwera mtengo. Pankhaniyi, sungani deta mu njira ina yapadera.
Zopanda seva pazayankho za OLTP
Chitsanzo cham'mbuyomu chinayang'ana ntchito za OLAP (analytical). Tsopano tiyeni tiwone ntchito za OLTP.
Tiyeni tingoyerekeza scalable PostgreSQL kapena MySQL. Tiyeni tikweze chitsanzo chokhazikika cha PostgreSQL kapena MySQL yokhala ndi zinthu zochepa. Chitsanzochi chikalandira katundu wochulukirapo, tidzalumikiza zofananira zina zomwe tidzagawirako gawo lazowerengera. Ngati palibe zopempha kapena katundu, timazimitsa zojambulazo. Chitsanzo choyamba ndi mbuye, ndipo ena onse ndi ofanizira.
Lingaliro ili likugwiritsidwa ntchito mu database yotchedwa Aurora Serverless AWS. Mfundoyi ndi yophweka: zopempha kuchokera kuzinthu zakunja zimavomerezedwa ndi zombo za proxy. Kuwona kuchuluka kwa katunduyo, kumagawa zida zamakompyuta kuchokera kunthawi yocheperako - kulumikizana kumapangidwa mwachangu momwe kungathekere. Kuyimitsa zochitika kumachitika chimodzimodzi.
Mkati mwa Aurora pali lingaliro la Aurora Capacity Unit, ACU. Ichi ndi (moyenera) chitsanzo (seva). ACU iliyonse yeniyeni ikhoza kukhala mbuye kapena kapolo. Chigawo chilichonse cha Mphamvu chili ndi RAM yake, purosesa ndi disk yaying'ono. Chifukwa chake, m'modzi ndi mbuye, ena onse amawerengedwa mongoyerekeza.
Chiwerengero cha ma Aurora Capacity Units omwe akuyenda ndi gawo losinthika. Kuchuluka kocheperako kumatha kukhala chimodzi kapena ziro (panthawiyi, database sikugwira ntchito ngati palibe zopempha).
Maziko akalandira zopempha, zombo zofananira zimakweza Aurora CapacityUnits, ndikuwonjezera magwiridwe antchito adongosolo. Kutha kuonjezera ndi kuchepetsa zothandizira kumapangitsa kuti dongosololi "ligwedezeke" zothandizira: kuwonetseratu ma ACUs (kuwasintha ndi atsopano) ndikutulutsa zosintha zonse zomwe zachotsedwa.
Maziko a Aurora Serverless amatha kukulitsa kuchuluka kwa kuwerenga. Koma zolembedwa sizikunena izi mwachindunji. Zingamveke ngati atha kukweza mbuye wambiri. Palibe zamatsenga.
Nawonsonkhokwe iyi ndiyoyenera kupewa kuwononga ndalama zambiri pamakina omwe ali ndi mwayi wosadziwika bwino. Mwachitsanzo, popanga MVP kapena makhadi otsatsa malonda, nthawi zambiri sitimayembekezera katundu wokhazikika. Chifukwa chake, ngati palibe mwayi wopeza, sitilipira milandu. Pamene katundu wosayembekezereka achitika, mwachitsanzo pambuyo pa msonkhano kapena ntchito yotsatsa malonda, makamu a anthu amayendera malowa ndipo katunduyo amawonjezeka kwambiri, Aurora Serverless imangotenga katunduyo ndikugwirizanitsa mwamsanga zinthu zomwe zikusowa (ACU). Kenako msonkhano umadutsa, aliyense amaiwala za mawonekedwe, ma seva (ACU) amapita mdima, ndipo mtengo umatsika mpaka zero - yabwino.
Yankho ili siloyenera kukweza mokhazikika chifukwa silimakulitsa zolembazo. Malumikizidwe onsewa ndi kuchotsedwa kwazinthu kumachitika pamalo otchedwa "scale point" - nthawi yomwe database siyimathandizidwa ndi zochitika kapena matebulo osakhalitsa. Mwachitsanzo, pasanathe sabata gawo la sikelo silingachitike, ndipo mazikowo amagwira ntchito pazida zomwezo ndipo sangathe kukulitsa kapena kupikisana.
Palibe zamatsenga - ndizokhazikika PostgreSQL. Koma njira yowonjezerera makina ndi kuwachotsa ndi yamagetsi.
Zopanda seva ndi mapangidwe
Aurora Serverless ndi nkhokwe yakale yolembedwanso kuti mtambo ugwiritse ntchito zina mwazabwino za Serverless. Ndipo tsopano ndikuwuzani za maziko, omwe poyamba adalembedwera mtambo, chifukwa cha njira yopanda seva - Seva-by-design. Idapangidwa nthawi yomweyo popanda kuganiza kuti idzayenda pa ma seva akuthupi.
Maziko awa amatchedwa Snowflake. Ili ndi makiyi atatu.
Choyamba ndi metadata block. Iyi ndi ntchito yokumbukira mwachangu yomwe imathetsa zovuta ndi chitetezo, metadata, zochitika, ndi kukhathamiritsa kwamafunso (zowonetsedwa pachithunzi chakumanzere).
Chida chachiwiri ndi gulu lamagulu apakompyuta owerengera (m'fanizoli pali magulu abuluu).
Chida chachitatu ndi njira yosungiramo deta yochokera ku S3. S3 ndi yosungirako zinthu zopanda malire mu AWS, ngati Dropbox yopanda malire yamabizinesi.
Tiyeni tiwone momwe Snowflake imagwirira ntchito, pongoyambira kuzizira. Ndiko kuti, pali nkhokwe, deta imayikidwamo, palibe mafunso othamanga. Chifukwa chake, ngati palibe zopempha ku nkhokwe, ndiye kuti takweza metadata yofulumira mu-memory (block yoyamba). Ndipo tili ndi S3 yosungirako, komwe deta ya tebulo imasungidwa, yogawidwa mu otchedwa micropartitions. Kuti zikhale zosavuta: ngati tebulo lili ndi zochitika, ndiye kuti ma micropartitions ndi masiku ochita. Tsiku lililonse ndi gawo lapadera la micropartition, fayilo yosiyana. Ndipo pamene database ikugwira ntchito motere, mumangolipira malo omwe ali ndi deta. Komanso, mlingo pa mpando ndi wotsika kwambiri (makamaka poganizira kupanikizika kwakukulu). Ntchito ya metadata imagwiranso ntchito nthawi zonse, koma simufunika zida zambiri kuti mukwaniritse mafunso, ndipo ntchitoyo imatha kuonedwa ngati shareware.
Tsopano tiyerekeze kuti wosuta adabwera ku database yathu ndikutumiza funso la SQL. Funso la SQL limatumizidwa nthawi yomweyo ku Metadata kuti ikonzedwe. Chifukwa chake, polandira pempho, ntchitoyi imasanthula pempho, zomwe zilipo, zilolezo za ogwiritsa ntchito ndipo, ngati zonse zili bwino, imapanga dongosolo lokonzekera pempholo.
Pambuyo pake, ntchitoyi imayambitsa kukhazikitsidwa kwa gulu la makompyuta. A computing cluster ndi gulu la ma seva omwe amawerengera. Ndiye kuti, ili ndi gulu lomwe limatha kukhala ndi seva 1, ma seva 2, 4, 8, 16, 32 - ochuluka momwe mukufunira. Mumaponya pempho ndipo kukhazikitsidwa kwa gululi nthawi yomweyo kumayamba. Zimatengadi masekondi.
Kenako, gululi litayamba, ma micropartitions omwe amafunikira kuti agwiritse ntchito pempho lanu amayamba kukopera mgulu kuchokera ku S3. Ndiye kuti, tiyerekeze kuti kuti mufunse funso la SQL mufunika magawo awiri patebulo limodzi ndi limodzi kuchokera pachiwiri. Pachifukwa ichi, magawo atatu okha ndi omwe adzakoperedwe kumagulu, osati matebulo onse. Ichi ndichifukwa chake, ndipo ndendende chifukwa chilichonse chili mkati mwa malo amodzi a data ndikulumikizidwa ndi mayendedwe othamanga kwambiri, njira yonse yosinthira imachitika mwachangu kwambiri: m'masekondi, kawirikawiri m'mphindi, pokhapokha tikulankhula za zopempha zowopsa. Chifukwa chake, ma micropartitions amakopera kugulu lamakompyuta, ndipo, mukamaliza, funso la SQL limaperekedwa pagulu la makompyuta ili. Chotsatira cha pempholi chikhoza kukhala mzere umodzi, mizere ingapo kapena tebulo - zimatumizidwa kunja kwa wogwiritsa ntchito kuti azitha kuzitsitsa, kuziwonetsera mu chida chake cha BI, kapena kuchigwiritsa ntchito mwanjira ina.
Funso lililonse la SQL silingangowerenga zophatikizika kuchokera pazomwe zidatsitsidwa kale, komanso kutsitsa / kupanga zatsopano mumsungidwe. Ndiko kuti, likhoza kukhala funso lomwe, mwachitsanzo, limayika zolemba zatsopano patebulo lina, zomwe zimatsogolera ku mawonekedwe a gawo latsopano pamagulu apakompyuta, omwe, nawonso, amangosungidwa mu yosungirako imodzi ya S3.
Zomwe tafotokozazi, kuyambira pakufika kwa wogwiritsa ntchito mpaka pakukweza gulu, kutsitsa deta, kufunsa mafunso, kupeza zotsatira, kumalipidwa pamlingo wa mphindi zogwiritsira ntchito gulu lokwezeka la makompyuta, nyumba yosungiramo zinthu zakale. Mtengo umasiyanasiyana kutengera dera la AWS ndi kukula kwamagulu, koma pafupifupi ndi madola angapo pa ola limodzi. Gulu la makina anayi ndi lokwera mtengo kuwirikiza kawiri kuposa gulu la makina awiri, ndipo gulu la makina asanu ndi atatu ndilokwera mtengo kuwirikiza kawiri. Zosankha za 16, makina a 32 alipo, kutengera zovuta za zopempha. Koma mumalipira mphindi zokhazo pamene gulu likuyenda, chifukwa ngati palibe zopempha, mumachotsa manja anu, ndipo mutatha mphindi 5-10 mukudikirira (parameter yosinthika) idzatuluka yokha, masulani zida ndikukhala mfulu.
Chochitika chenichenicho ndi pamene mutumiza pempho, gululo limatuluka, kunena kuti, mu miniti, imawerengera mphindi ina, ndiye mphindi zisanu kuti mutseke, ndipo mumathera kulipira kwa mphindi zisanu ndi ziwiri za ntchito ya gululi, ndipo osati kwa miyezi ndi zaka.
Chochitika choyamba chofotokozedwa chogwiritsa ntchito Snowflake pakugwiritsa ntchito munthu m'modzi. Tsopano tiyeni tiyerekeze kuti pali ogwiritsa ntchito ambiri, omwe ali pafupi ndi zochitika zenizeni.
Tinene kuti tili ndi owunika ambiri komanso malipoti a Tableau omwe amangokhalira kuwononga database yathu ndi mafunso ambiri osavuta a SQL.
Kuphatikiza apo, tinene kuti tili ndi asayansi a Data omwe akuyesera kuchita zinthu zoopsa kwambiri ndi data, amagwira ntchito ndi ma Terabytes makumi, kusanthula mabiliyoni ndi mathililiyoni a mizere ya data.
Pamitundu iwiri yantchito yomwe yafotokozedwa pamwambapa, Snowflake imakupatsani mwayi wokweza magulu angapo odziyimira pawokha amitundu yosiyanasiyana. Kuphatikiza apo, magulu apakompyutawa amagwira ntchito paokha, koma ndi data yofananira.
Pamafunso ambiri opepuka, mutha kukweza timagulu ting'onoting'ono 2-3, pafupifupi makina awiri lililonse. Izi zitha kuchitika, mwa zina, pogwiritsa ntchito zosintha zokha. Kotero inu mukuti, βChipale chofewa, kwezani kagulu kakang'ono. Ngati katunduyo akuwonjezeka pamwamba pa chizindikiro china, kwezani sekondi yofanana, yachitatu. Pamene katunduyo ayamba kuchepa, zimitsani zochulukirapo. " Kotero kuti ngakhale ofufuza angati abwere ndikuyamba kuyang'ana malipoti, aliyense ali ndi zothandizira zokwanira.
Panthawi imodzimodziyo, ngati ofufuza akugona ndipo palibe amene akuyang'ana malipoti, masango amatha kukhala mdima, ndipo mumasiya kuwalipira.
Nthawi yomweyo, pamafunso olemetsa (kuchokera ku Data Scientists), mutha kukweza gulu limodzi lalikulu kwambiri pamakina 32. Tsango ili lidzalipidwanso mphindi ndi maola amenewo pomwe pempho lanu lalikulu likuyenda pamenepo.
Mwayi womwe wafotokozedwa pamwambapa umakupatsani mwayi wogawa osati 2 yokha, komanso mitundu yambiri yantchito m'magulu (ETL, kuyang'anira, kupanga lipoti, ...).
Tiyeni tifotokoze mwachidule Snowflake. Maziko amaphatikiza lingaliro lokongola komanso kukhazikitsa koyenera. Ku ManyChat, timagwiritsa ntchito Snowflake kusanthula zonse zomwe tili nazo. Tilibe masango atatu, monga chitsanzo, koma kuyambira 5 mpaka 9, amitundu yosiyanasiyana. Tili ndi makina 16 wamba, makina awiri, komanso makina ang'onoang'ono a 2 pa ntchito zina. Amagawa bwino katunduyo ndipo amatilola kuti tipulumutse zambiri.
Dongosolo lankhokwe limakulitsa bwino kuchuluka kwa kuwerenga ndi kulemba. Uku ndi kusiyana kwakukulu komanso kupambana kwakukulu poyerekeza ndi "Aurora" yomweyi, yomwe inkangonyamula katundu wowerenga. Snowflake imakupatsani mwayi wowonjezera ntchito yanu yolemba ndi magulu apakompyuta awa. Izi ndizo, monga ndanenera, timagwiritsa ntchito magulu angapo mu ManyChat, magulu ang'onoang'ono ndi ang'onoang'ono amagwiritsidwa ntchito makamaka ku ETL, pokweza deta. Ndipo akatswiri amakhala kale pamagulu apakati, omwe sanakhudzidwe ndi katundu wa ETL, choncho amagwira ntchito mofulumira kwambiri.
Chifukwa chake, nkhokweyo ndiyoyenera ntchito za OLAP. Komabe, mwatsoka, sikugwira ntchito pazantchito za OLTP. Choyamba, database iyi ndi columnar, ndi zotsatira zake zonse. Kachiwiri, njira yokhayo, pamene pempho lililonse, ngati kuli kofunikira, mumakweza gulu la makompyuta ndikulidzaza ndi deta, mwatsoka, silinafikebe mokwanira pa katundu wa OLTP. Kudikirira masekondi pa ntchito za OLAP ndikwabwinobwino, koma kwa ntchito za OLTP ndikosavomerezeka; 100 ms ingakhale yabwinoko, kapena 10 ms ingakhale yabwinoko.
Zotsatira
Dongosolo lopanda seva ndilotheka pogawa nkhokwe mu magawo Osavomerezeka ndi Okhazikika. Mwina mwazindikira kuti m'zitsanzo zonse zomwe zili pamwambapa, gawo la Stateful ndiloti, kusungirako magawo ang'onoang'ono mu S3, ndipo Stateless ndiye optimizer, akugwira ntchito ndi metadata, kuthana ndi nkhani zachitetezo zomwe zitha kufotokozedwa ngati ntchito zodziyimira pawokha zopanda Stateless.
Kuyankha mafunso a SQL kumatha kuwonedwanso ngati ntchito zopepuka zomwe zimatha kuwonekera mopanda seva, monga magulu a makompyuta a Snowflake, kutsitsa zofunikira zokha, funsanizo ndi "kutuluka."
Zosungira zopanga zopanda seva zilipo kale kuti zigwiritsidwe ntchito, zikugwira ntchito. Zosungira zopanda sevazi zakonzeka kale kugwira ntchito za OLAP. Tsoka ilo, pa ntchito za OLTP amagwiritsidwa ntchito ... ndi ma nuances, popeza pali malire. Kumbali imodzi, uku ndi kuchotsera. Koma, kumbali ina, uwu ndi mwayi. Mwina m'modzi mwa owerenga apeza njira yopangira database ya OLTP kukhala yopanda seva, popanda malire a Aurora.
Ndikukhulupirira kuti mwapeza zosangalatsa. Serverless ndi tsogolo :)
Source: www.habr.com