Mhoroi mose! Ini ndinonzi Golov Nikolay. Pakutanga, ndakashanda kuAvito uye ndakatarisira Data Platform kwemakore matanhatu, kureva kuti, ndakashanda pane zvose zvinyorwa: analytical (Vertica, ClickHouse), kutenderera uye OLTP (Redis, Tarantool, VoltDB, MongoDB, PostgreSQL). Munguva iyi, ndakabata nenhamba huru yedatabase - yakasiyana chaizvo uye isina kujairika, uye neasina-standard kesi dzekushandisa kwavo.
Iye zvino ndiri kushanda paManyChat. Muchidimbu, uku ndiko kutanga - kutsva, kuda chinzvimbo uye kukura nekukurumidza. Uye pandakatanga kujoinha kambani, mubvunzo wekare wakamuka: "Chii chinofanira kutora mudiki kubva kuDBMS nemusika wedhatabhesi?"
Muchinyorwa chino, zvichibva pamushumo wangu pa , ndichapindura mubvunzo uyu. Vhidhiyo vhezheni yemushumo inowanikwa pa .
Anowanzo zivikanwa dhatabhesi 2020
Ndi2020, ndakatarisa-tarisa ndikaona mhando nhatu dzedhatabhesi.
First type - classic OLTP dhatabhesi: PostgreSQL, SQL Server, Oracle, MySQL. Iwo akanyorwa kare kare, asi achiri akakosha nekuti akajairwa kunharaunda yekuvandudza.
Rudzi rwechipiri ndirwo mabhesi kubva ku "zero". Vakaedza kubva pamaitiro echinyakare nekusiya SQL, zvimiro zvechinyakare uye ACID, nekuwedzera yakavakirwa-mukati sharding uye zvimwe zvinokwezva maficha. Semuenzaniso, iyi iCassandra, MongoDB, Redis kana Tarantool. Ese mhinduro idzi dzaida kupa musika chimwe chinhu chakanyanya chitsva uye chakatora niche yavo nekuti zvakazove zvakanyanyonakira kune mamwe mabasa. Ini ndicharatidza aya dhatabhesi ane amburera izwi NOSQL.
Iwo "zero" apera, takajaira NOSQL dhatabhesi, uye nyika, kubva pamaonero angu, yakatora danho rinotevera - kusvika yakachengetedzwa databases. Aya dhatabhesi ane musimboti wakafanana neyekare OLTP dhatabhesi kana matsva eNoSQL. Asi ivo havana kudiwa kweDBA neDevOps uye vanomhanya pane inogadziriswa hardware mumakore. Kune anovandudza, ichi "chingori chigadziko" chinoshanda pane imwe nzvimbo, asi hapana ane hanya kuti chakaiswa sei pavhavha, uyo akagadzirisa sevha uye anoivandudza.
Mienzaniso yemadhatabhesi akadaro:
- AWS RDS ndeye inogadziriswa wrapper yePostgreSQL/MySQL.
- DynamoDB ndeye AWS analogue yegwaro rakavakirwa dhatabhesi, rakafanana neRedis neMongoDB.
- Amazon Redshift inzvimbo yakachengetedzwa yekuongorora dhatabhesi.
Aya ndiwo ekare dhatabhesi, asi akasimudzwa munzvimbo inochengetedzwa, pasina chikonzero chekushanda nehardware.
Cherechedza. Mienzaniso inotorwa yenzvimbo yeAWS, asi analogues avo aripowo muMicrosoft Azure, Google Cloud, kana Yandex.Cloud.
Chii chitsva pamusoro peizvi? Muna 2020, hapana chimwe cheizvi.
Serverless pfungwa
Chii chaizvo chitsva pamusika muna 2020 haina serverless kana serverless mhinduro.
Ini ndichaedza kutsanangura kuti izvi zvinorevei kushandisa muenzaniso wenguva dzose sevhisi kana backend application.
Kuti titumire akajairika backend application, tinotenga kana kurenda sevha, kukopa kodhi pairi, kushambadza iyo yekupedzisira kunze uye kugara tichibhadhara rendi, magetsi uye data center services. Iyi ndiyo standard scheme.
Pane imwe nzira here? Ne serverless masevhisi unogona.
Chii chiri kutariswa kweiyi nzira: hapana sevha, hapana kana kurenda chiitiko chaicho mugore. Kuti uendese sevhisi, kopira iyo kodhi (mafunctions) kune repository uye ishambadzire kusvika kumagumo. Ipapo isu tinongobhadhara kune imwe neimwe kufona kune iyi basa, tichiregeredza zvachose iyo Hardware kwainotemerwa.
Ini ndichaedza kuenzanisira nzira iyi nemifananidzo.
Classic deployment. Tine sevhisi ine mumwe mutoro. Isu tinosimudza zviitiko zviviri: maseva emuviri kana zviitiko muAWS. Zvikumbiro zvekunze zvinotumirwa kune zviitiko izvi uye zvinogadziriswa ipapo.
Sezvauri kuona pamufananidzo, maseva haana kuraswa zvakaenzana. Imwe ndeye 100% inoshandiswa, pane zvikumbiro zviviri, uye imwe inongova 50% - isina kuita zvishoma. Kana zvisiri zvikumbiro zvitatu zvinosvika, asi makumi matatu, ipapo iyo yose system haizokwanisi kutarisana nemutoro uye inotanga kuderera.
Serverless deployment. Munzvimbo isina server, sevhisi yakadai haina zviitiko kana maseva. Kune imwe dziva rezviwanikwa zvinopisa - zvidiki zvakagadzirirwa Docker midziyo ine deployed function code. Iyo sisitimu inogashira zvikumbiro zvekunze uye kune yega yega iyo serverless sisitimu inosimudza diki mudziyo ine kodhi: inogadzirisa ichi chikumbiro uye inouraya mudziyo.
Chikumbiro chimwe - mudziyo mumwe wakasimudzwa, 1000 zvikumbiro - 1000 midziyo. Uye kutumirwa pamaseva ehardware ratova basa remupi wegore. Iyo yakanyatsovanzwa ne serverless framework. Mune iyi pfungwa tinobhadhara kune yese kufona. Semuenzaniso, kufona kumwe kwakauya pazuva - takabhadhara kufona kumwe, miriyoni yakauya paminiti - takabhadhara miriyoni. Kana kuti mune yechipiri, izvi zvinoitikawo.
Iyo pfungwa yekuburitsa serverless basa yakakodzera kune isina nyika sevhisi. Uye kana iwe uchida (nyika) statefull sevhisi, saka isu tinowedzera dhatabhesi kune sevhisi. Muchiitiko ichi, kana zvasvika pakushanda nehurumende, imwe neimwe statefull basa rinongonyora uye kuverenga kubva mudhatabhesi. Uyezve, kubva mudura reimwe remhando nhatu dzakatsanangurwa pakutanga kwechinyorwa.
Ndeipi mipimo yakajairika yemadatabase ese aya? Iyi ndiyo mitengo yegore rinogara richishandiswa kana hardware server (kana akati wandei maseva). Izvo hazvina basa kuti isu tinoshandisa yekare kana yakachengetedzwa dhatabhesi, kunyangwe isu tine maDevops uye admin kana kwete, isu tichiri kubhadhara Hardware, magetsi uye data centre yekurenda 24/7. Kana isu tine classic base, tinobhadhara tenzi nemuranda. Kana iri dhatabhesi rakaremerwa zvakanyanya, tinobhadhara gumi, makumi maviri kana makumi matatu maseva, uye tinobhadhara nguva dzose.
Kuvapo kwemaseva akachengetwa zvachose muchimiro chemutengo kwaimboonekwa sehuipi hunodiwa. Madhijitari akajairwa anewo mamwe matambudziko, senge miganho pahuwandu hwekubatanidza, kuyera zvirambidzo, geo-yakagoverwa kubvumirana - ivo vanogona neimwe nzira kugadziriswa mune mamwe dhatabhesi, asi kwete zvese kamwechete uye kwete zvakanakira.
Serverless database - dzidziso
Mubvunzo wa2020: zvinoita here kuita dhatabhesi serverless futi? Wese munhu akanzwa nezve serverless backend ... ngatiedzei kuita dhatabhesi serverless?
Izvi zvinonzwika zvisinganzwisisike, nekuti dhatabhesi ibasa rakazara, risina kunyatsokodzera serverless zvivakwa. Panguva imwecheteyo, mamiriro e database yakakura kwazvo: gigabytes, terabytes, uye mune analytical databases kunyange petabytes. Hazvisi nyore kuisimudza mumidziyo yakareruka yeDocker.
Kune rimwe divi, angangoita ese emazuva ano dhatabhesi ane hukuru hukuru hwepfungwa uye zvikamu: kutengeserana, kuvimbika kurongeka, maitiro, kutsamira kwehukama uye yakawanda pfungwa. Kune yakawanda yakawanda yedatabase logic, diki nyika inokwana. Gigabytes uye Terabytes anoshandiswa zvakananga nechikamu chidiki chete cheiyo database logic inobatanidzwa mukuita zvakananga mibvunzo.
Saizvozvo, zano nderekuti: kana chikamu chepfungwa chichibvumira kuurayiwa kusingaverengeki, wadii kupatsanura hwaro kuita zvikamu zveStateful uye Stateless.
Serverless kune OLAP mhinduro
Ngationei kuti kucheka dhatabhesi muzvikamu zveStateful uye Stateless zvingaite sekushandisa mienzaniso inoshanda.
Semuenzaniso, isu tine analytical database: data rekunze (cylinder tsvuku kuruboshwe), chirongwa cheETL chinotakura data mudhatabhesi, uye muongorori anotumira mibvunzo yeSQL kudhatabhesi. Iyi ndiyo yekare data warehouse operation scheme.
Muchirongwa ichi, ETL inoitwa kamwe chete. Ipapo iwe unofanirwa kugara uchibhadharira maseva pane iyo dhatabhesi inomhanya nedata rakazadzwa neETL, kuitira kuti pane chimwe chinhu chekutumira mibvunzo.
Ngatitarisei imwe nzira inoshandiswa muAWS Athena Serverless. Iko hakuna hardware yakatsaurirwa zvachose inochengetwa data yakatorwa. Panzvimbo peizvi:
- Mushandisi anotumira mubvunzo weSQL kuAthena. Iyo Athena optimizer inoongorora iyo SQL query uye inotsvaga metadata chitoro (Metadata) kune chaiyo data inodiwa kuita mubvunzo.
- Iyo optimizer, yakavakirwa pane yakaunganidzwa data, inodhawunirodha data inodiwa kubva kune ekunze masosi mukuchengetedza kwenguva pfupi (yenguva dhatabhesi).
- Mubvunzo weSQL kubva kumushandisi unoitwa mukuchengetedza kwenguva pfupi uye mhedzisiro inodzoserwa kumushandisi.
- Kuchengetedza kwenguva pfupi kunocheneswa uye zviwanikwa zvinoburitswa.
Muchivakwa ichi, isu tinongobhadhara chete maitiro ekuita chikumbiro. Hapana zvikumbiro - hapana mari.
Iyi inzira yekushanda uye inoshandiswa kwete muAthena Serverless chete, asiwo muRedshift Spectrum (muAWS).
Muenzaniso weAthena unoratidza kuti Serverless dhatabhesi inoshanda pamibvunzo chaiyo ine makumi nemazana eTerabytes yedata. Mazana eTerabytes achada mazana emaseva, asi isu hatifanirwe kuvabhadhara - tinobhadhara zvikumbiro. Iko kumhanya kwechikumbiro chega chega (kwakanyanya) kuderera kana ichienzaniswa nehunyanzvi hwekuongorora dhatabhesi seVertica, asi isu hatibhadhare nguva yekudzikira.
Dhatabhesi rakadaro rinoshanda kune isingawanzo kuongorora ad-hoc mibvunzo. Semuyenzaniso, kana isu tikangoerekana tafunga kuyedza fungidziro pane imwe gigantic yakawanda yedata. Athena yakakwana kune idzi kesi. Pakukumbira nguva dzose, hurongwa hwakadaro hunodhura. Mune ino kesi, cache iyo data mune imwe yakasarudzika mhinduro.
Serverless yeOLTP mhinduro
Muenzaniso wekare wakatarisa OLAP (analytical) mabasa. Zvino ngatitarisei mabasa eOLTP.
Ngatimbofungidzira scalable PostgreSQL kana MySQL. Ngatisimudzei yakajairwa inogadziriswa muenzaniso PostgreSQL kana MySQL ine zviwanikwa zvishoma. Kana iyo muenzaniso inogamuchira yakawanda mutoro, isu tichabatanidza mamwe makopi ayo isu tichagovera chikamu chekuverenga mutoro. Kana pasina zvikumbiro kana mutoro, tinodzima replicas. Muenzaniso wekutanga ndewe tenzi, uye vamwe vese vakafanana.
Pfungwa iyi inoshandiswa mudhatabhesi inonzi Aurora Serverless AWS. Nheyo yacho iri nyore: zvikumbiro zvekunze zvinogamuchirwa neproxy fleet. Kuona kuwanda kwemutoro, inogovera zviwanikwa zvekombuta kubva kusati-kwadziya zvishoma zviitiko - kubatana kunoitwa nekukurumidza sezvinobvira. Kuremara zviitiko kunoitika nenzira imwecheteyo.
Mukati meAurora mune pfungwa yeAurora Capacity Unit, ACU. Izvi (zvichienderana) muenzaniso (server). Imwe neimwe chaiyo ACU inogona kuve tenzi kana muranda. Imwe neimwe Capacity Unit ine yayo RAM, processor uye shoma dhisiki. Saizvozvo, mumwe ndiye tenzi, vamwe vese vanoverengerwa chete replicas.
Huwandu hweaya maAurora Capacity Units anomhanya ndeye inogadziriswa parameter. Nhamba shoma inogona kuva imwe kana zero (munyaya iyi, database haishande kana pasina zvikumbiro).
Kana hwaro hwagamuchira zvikumbiro, iyo proxy ngarava inosimudza Aurora CapacityUnits, ichiwedzera masisitimu ekuita zviwanikwa. Iko kugona kuwedzera uye kuderedza zviwanikwa kunobvumira sisitimu "juggle" zviwanikwa: kuratidza otomatiki maACU (achitsiva iwo matsva) uye kuburitsa zvese zvazvino zvigadziriso kune zvakabviswa zviwanikwa.
Iyo Aurora Serverless base inogona kuyera mutoro wekuverenga. Asi magwaro haatauri izvi zvakananga. Inogona kunzwa sekunge vanogona kusimudza akawanda-tenzi. Hakuna mashiripiti.
Iyi dhatabhesi yakanyatsokodzera kudzivirira kushandisa mari hombe pane masisitimu ane kusingafungidzirwe kuwana. Semuenzaniso, pakugadzira MVP kana kushambadzira bhizinesi makadhi mawebhusaiti, isu kazhinji hatitarisiri mutoro wakagadzikana. Saizvozvo, kana pasina mukana, isu hatibhadharire zviitiko. Kana mutoro usingatarisirwi ukaitika, semuenzaniso mushure memusangano kana mushandirapamwe wekushambadzira, mhomho yevanhu inoshanyira saiti uye mutoro unowedzera zvakanyanya, Aurora Serverless inongotora mutoro uyu uye nekukurumidza inobatanidza zvisipo (ACU). Ipapo musangano unopfuura, munhu wese anokanganwa nezve prototype, maseva (ACU) anosviba, uye mutengo unodonha kusvika zero - zviri nyore.
Iyi mhinduro haina kukodzera kukwirisa yakagadzikana nekuti haina kuyera mutoro wekunyora. Zvese izvi zvinongedzo uye kubviswa kwezviwanikwa kunoitika pane inonzi "scale point" - imwe nguva iyo dhatabhesi isingatsigirwe nekutengeserana kana matafura enguva pfupi. Semuenzaniso, mukati mevhiki chiyero chechiyero chinogona kusaitika, uye hwaro hunoshanda pane imwechete zviwanikwa uye haigone kuwedzera kana chibvumirano.
Iko hakuna mashiripiti - ndeyenguva dzose PostgreSQL. Asi maitiro ekuwedzera michina uye kuibvisa ine chikamu otomatiki.
Serverless nekugadzira
Aurora Serverless ndeyekare dhatabhesi yakanyorwa zvakare kuti gore ritore mukana wemamwe mabhenefiti eServerless. Uye zvino ini ndichakuudza nezve base, iyo yakatanga kunyorerwa gore, kune serverless nzira - Serverless-ne-dhizaini. Yakagadzirwa pakarepo pasina fungidziro yekuti yaizomhanya pamaseva emuviri.
Iyi nheyo inonzi Snowflake. Iine mabhuroko matatu akakosha.
Yekutanga ndeye metadata block. Iyi inokurumidza-mu-memory sevhisi inogadzirisa nyaya nechengetedzo, metadata, kutengeserana, uye optimization yemibvunzo (inoratidzwa mumufananidzo uri kuruboshwe).
Yechipiri bhuroka seti yemadhizaini emakomputa ekuverenga (mumufananidzo pane seti yebhuruu madenderedzwa).
Yechitatu block ndeye data yekuchengetedza system yakavakirwa paS3. S3 haina dimensionless chinhu kuchengetedza muAWS, senge isina dimensionless Dropbox yebhizinesi.
Ngationei kuti Snowflake inoshanda sei, tichifunga kutanga kutonhora. Ndiko kuti, pane dhatabhesi, iyo data inotakurwa mairi, hapana inomhanya mibvunzo. Saizvozvo, kana pasina zvikumbiro kune dhatabhesi, saka isu takasimudza nekukurumidza mu-memory Metadata sevhisi (yekutanga block). Uye isu tine S3 kuchengetedza, uko data retafura rinochengetwa, rakakamurwa kuita zvinonzi micropartitions. Kuti zvive nyore: kana tafura iine kutengeserana, saka micropartitions ndiwo mazuva ekutengeserana. Zuva rega rega i micropartition yakasiyana, faira yakasiyana. Uye kana dhatabhesi ichishanda mune iyi modhi, iwe unongobhadhara chete nzvimbo inogarwa nedata. Uyezve, chiyero chechigaro chakanyanya kuderera (kunyanya tichifunga nezvekumanikidza kwakakosha). Iyo metadata sevhisi inoshandawo nguva dzose, asi iwe haudi yakawanda zviwanikwa kuti ugone kukwirisa mibvunzo, uye sevhisi inogona kunzi shareware.
Zvino ngatimbofungidzira kuti mushandisi akauya kune yedu database uye akatumira mubvunzo weSQL. Mubvunzo weSQL unobva watumirwa kuMetadata sevhisi kuti igadziriswe. Saizvozvo, pakugamuchira chikumbiro, sevhisi iyi inoongorora chikumbiro, data iripo, mvumo yemushandisi uye, kana zvese zvakanaka, inogadzira chirongwa chekugadzirisa chikumbiro.
Tevere, sevhisi inotanga kutangwa kweiyo computing cluster. A computing cluster iboka remaseva anoita masvomhu. Kureva kuti, iri isumbu rinogona kunge riine 1 sevha, 2 maseva, 4, 8, 16, 32 - akawanda sezvaunoda. Iwe unokanda chikumbiro uye kutangwa kwechikwata ichi kunobva kwatanga. Zvinotora masekonzi.
Tevere, mushure mekunge cluster yatanga, iwo mamicropartitions anodiwa kugadzirisa chikumbiro chako anotanga kukopwa musumbu kubva kuS3. Ndokunge, ngatimbofungidzira kuti kuita SQL mubvunzo unoda zvikamu zviviri kubva patafura imwe uye imwe kubva yechipiri. Muchiitiko ichi, zvikamu zvitatu chete zvinodiwa zvichakopwa kune sumbu, uye kwete matafura ose zvachose. Ndokusaka, uye chaizvo nekuti zvese zviri mukati meimwe data data uye zvakabatanidzwa neanokurumidza chiteshi, iyo yese yekufambisa maitiro inoitika nekukurumidza: mumasekondi, kashoma mumaminitsi, kunze kwekunge tiri kutaura nezve zvimwe zvikumbiro zvinotyisa. Saizvozvo, ma micropartitions anoteedzerwa kucomputing cluster, uye, kana apedza, iyo SQL query inoitwa pane ino computing cluster. Mhedzisiro yechikumbiro ichi inogona kuve mutsara mumwe, mitsetse yakati wandei kana tafura - inotumirwa kunze kune mushandisi kuti agone kuidhawunirodha, kuiratidza muBI chishandiso chake, kana kuishandisa neimwe nzira.
Mubvunzo wega wega weSQL haugone kungoverenga zvakaunganidzwa kubva kune data rakamboremerwa, asiwo kurodha / kugadzira data nyowani mudhatabhesi. Ndiko kuti, inogona kunge iri mubvunzo wekuti, semuenzaniso, inoisa zvinyorwa zvitsva mune imwe tafura, izvo zvinotungamira pakuonekwa kwechikamu chitsva pane computing cluster, iyo, zvakare, inochengetwa otomatiki mune imwechete S3 yekuchengetedza.
Mamiriro ezvinhu anotsanangurwa pamusoro apa, kubva pakusvika kwemushandisi kusvika pakusimudzwa kweboka, kurodha data, kuita mibvunzo, kuwana mhinduro, inobhadharwa pamwero wemaminetsi ekushandisa yakasimudzwa virtual computing cluster, virtual warehouse. Mutengo unosiyana zvichienderana nenzvimbo yeAWS uye saizi yesumbu, asi paavhareji imadhora mashoma paawa. Boka remichina mina rinodhura zvakapetwa kaviri sesumbu remichina miviri, uye sumbu remichina misere richiri kudhura zvakapetwa kaviri. Sarudzo dze16, 32 michina iripo, zvichienderana nekuoma kwezvikumbiro. Asi iwe unobhadhara chete iwo maminetsi apo cluster iri kunyatso mhanya, nekuti kana pasina zvikumbiro, iwe unobvisa maoko ako, uye mushure me5-10 maminetsi ekumirira (inogadziriswa parameter) ichabuda yega, sunungura zviwanikwa uye uve wakasununguka.
Mamiriro ezvinhu anonyatsoitika ndeapo paunotumira chikumbiro, sumbu rinobuda, tichitaura, muminiti, rinoverenga imwe miniti, wozoita maminetsi mashanu kuvhara, uye unopedzisira wabhadhara maminetsi manomwe ekushanda kwesumbu iri, uye kwete kwemwedzi nemakore.
Mamiriro ekutanga anotsanangurwa uchishandisa Snowflake mune imwechete-mushandisi kuseta. Iye zvino ngatimbofungidzira kuti kune vakawanda vashandisi, iyo iri pedyo nemamiriro chaiwo.
Ngatitii isu tine vakawanda vanoongorora uye Tableau inoshuma iyo inogara ichibhowa dhatabhesi yedu nehuwandu hukuru hwemibvunzo yakapusa yekuongorora SQL.
Uye zvakare, ngatiti isu tine inventive Data Scientists vari kuyedza kuita zvinhu zvinotyisa nedata, vanoshanda nemakumi eTerabytes, vanoongorora mabhiriyoni nemabhiriyoni emitsara yedata.
Kune mhando mbiri dzebasa ratsanangurwa pamusoro apa, Snowflake inokutendera iwe kusimudza akati wandei akazvimirira makomputa masumbu eakasiyana masimba. Uyezve, aya masumbu emakomputa anoshanda akazvimirira, asi aine data rinoenderana.
Kune huwandu hukuru hwebvunzo dzakareruka, unogona kusimudza 2-3 masumbu madiki, angangoita 2 michina imwe neimwe. Maitiro aya anogona kuitwa, pakati pezvimwe zvinhu, uchishandisa otomatiki marongero. Saka unoti, βSnowflake, simudza kasumbu kadiki. Kana mutoro pairi ukawedzera pamusoro peimwe parameter, simudza yakafanana yechipiri, yechitatu. Kana mutoro ukatanga kudzikira, dzima zvakawandisa. β Zvekuti kunyangwe vaongorori vangani vanouya votanga kutarisa mareport, munhu wese ane zviwanikwa zvakakwana.
Panguva imwecheteyo, kana vaongorori vakarara uye pasina anotarisa mishumo, masumbu anogona kusviba zvachose, uye iwe unorega kuvabhadhara.
Panguva imwecheteyo, pamibvunzo inorema (kubva kuData Scientists), unogona kusimudza sumbu rakakura kwazvo remichina makumi matatu nemaviri. Iri sumbu rinozobhadharwa chete iwo maminetsi nemaawa apo chikumbiro chako chikuru chiri kushanda ipapo.
Mukana watsanangurwa pamusoro apa unobvumidza iwe kugovera kwete 2 chete, asiwo mamwe marudzi emutoro webasa mumasumbu (ETL, monitoring, report materialization,...).
Ngatipfupise Snowflake. Nheyo inobatanidza pfungwa yakanaka uye kushandiswa kunoshanda. PaManyChat, tinoshandisa Snowflake kuongorora data yose yatinayo. Isu hatina masumbu matatu, semuenzaniso, asi kubva pa5 kusvika ku9, ehukuru hwakasiyana. Isu tine zvakajairika 16-muchina, 2-muchina, uye zvakare super-diki 1-muchina mamwe mabasa. Vanobudirira kugovera mutoro uye vanotibvumira kuchengetedza zvakawanda.
Iyo dhatabhesi inobudirira kuyera kuverenga nekunyora mutoro. Uyu mutsauko wakakura uye budiriro huru ichienzaniswa neiyo "Aurora", iyo yaingotakura mutoro wekuverenga. Snowflake inokutendera iwe kuyera basa rako rekunyora neaya masumbu emakomputa. Ndiko, sezvandambotaura, tinoshandisa masumbu akati wandei muManyChat, madiki uye makuru-madiki masumbu anonyanya kushandiswa kuETL, kurodha data. Uye vaongorori vanotogara pamasumbu epakati, ayo asina kukanganiswa zvachose neETL mutoro, saka anoshanda nekukurumidza.
Saizvozvo, iyo dhatabhesi inonyatso kuenderana neOLAP mabasa. Nekudaro, zvinosuwisa, haisati yashanda kune OLTP mitoro yebasa. Chekutanga, iyi dhatabhesi ndeye columnar, ine zvese zvinozoitika. Chechipiri, iyo nzira pachayo, kana yechikumbiro chega chega, kana zvichidikanwa, iwe unosimudza sumbu rekombuta uye worifashura nedata, zvinosiririsa, haisati yakurumidza zvakakwana kune mitoro yeOLTP. Kumirira masekonzi emabasa eOLAP zvakajairika, asi kumabasa eOLTP hazvigamuchirwe; 100 ms zvingava nani, kana gumi zvingatove nani.
Mugumisiro
Iyo serverless dhatabhesi inogoneka nekugovanisa dhatabhesi kuita Stateless uye Stateful zvikamu. Iwe unogona kunge waona kuti mumienzaniso yese iri pamusoro, chikamu cheStateful ndechekuti, tichitaura, kuchengetedza madiki-zvikamu muS3, uye Stateless ndiyo inogadzirisa, ichishanda nemetadata, inobata nyaya dzekuchengetedza dzinogona kusimudzwa seyakazvimirira lightweight Stateless masevhisi.
Kuita SQL mibvunzo inogona zvakare kunzwisiswa seyakajeka-nyika masevhisi anogona kubuda mune serverless modhi, seSnowflake computing masumbu, dhawunirodha data rinodiwa chete, bvunza mubvunzo uye "buda."
Serverless yekugadzira level dhatabhesi dzatovepo kuti dzishandiswe, dziri kushanda. Aya maseva asina dhatabhesi akatogadzirira kubata OLAP mabasa. Zvinosuruvarisa, kumabasa eOLTP anoshandiswa ... ane nuances, sezvo pane zvisingakwanisi. Kune rimwe divi, iyi minus. Asi, kune rumwe rutivi, uyu mukana. Zvichida mumwe wevaverengi achawana nzira yekugadzira OLTP dhatabhesi isina serverless zvachose, pasina zvipimo zveAurora.
Ndinovimba wakazviwana zvichinakidza. Serverless ndiyo ramangwana :)
Source: www.habr.com