Depozita e p\u00ebrkohshme pastrohet, burimet lirohen.<\/li>\n<\/ul>\nN\u00eb k\u00ebt\u00eb arkitektur\u00eb, ne paguajm\u00eb vet\u00ebm p\u00ebr procesin e ekzekutimit t\u00eb pyetjes. Nuk ka pyetje \u2013 nuk ka shpenzime.<\/p>\n
N\u00eb shembullin e Athena, duket se databaza Serverless funksionon me k\u00ebrkesa reale me dhjet\u00ebra dhe qindra terabajt t\u00eb dh\u00ebnash. P\u00ebr qindra terabajt do t\u00eb nevojiten qindra server\u00eb, por ne nuk duhet t\u00eb paguajm\u00eb p\u00ebr ta \u2014 paguajm\u00eb vet\u00ebm p\u00ebr k\u00ebrkesat. Shpejt\u00ebsia e \u00e7do k\u00ebrkese \u00ebsht\u00eb (shum\u00eb) e ul\u00ebt n\u00eb krahasim me databazat analitike t\u00eb specializuara si Vertica, por nuk paguajm\u00eb p\u00ebr periudhat e papun\u00ebsis\u00eb.<\/p>\n
Kjo databaz\u00eb \u00ebsht\u00eb e aplikueshme p\u00ebr k\u00ebrkesat analitike ad-hoc t\u00eb rralla. P\u00ebr shembull, kur ne vendosim t\u00eb kontrollojm\u00eb nj\u00eb hipotez\u00eb mbi nj\u00eb sasi t\u00eb madhe t\u00eb dh\u00ebnash. P\u00ebr k\u00ebto raste, Athena \u00ebsht\u00eb ideale. P\u00ebr k\u00ebrkesat e rregullta, nj\u00eb sistem i till\u00eb b\u00ebhet i kushtuesh\u00ebm. N\u00eb k\u00ebt\u00eb rast, ruani t\u00eb dh\u00ebnat n\u00eb ndonj\u00eb zgjidhje t\u00eb specializuar. <\/p>\n
Serverless p\u00ebr zgjidhjet OLTP<\/h2>\n \nN\u00eb shembullin e m\u00ebparsh\u00ebm, u shqyrtuan detyra OLAP (analitike). Tani le t\u00eb shqyrtojm\u00eb detyra OLTP.<\/p>\n
Le t\u00eb p\u00ebrshkruajm\u00eb nj\u00eb PostgreSQL ose MySQL n\u00eb shkall\u00eb. Le t\u00eb ngrem\u00eb nj\u00eb instanc\u00eb t\u00eb menaxhuar PostgreSQL ose MySQL me burime minimale. Kur instanca t\u00eb ket\u00eb nj\u00eb ngarkes\u00eb m\u00eb t\u00eb madhe, do t\u00eb lidhim replikat shtes\u00eb, ku do t\u00eb shp\u00ebrndajm\u00eb nj\u00eb pjes\u00eb t\u00eb ngarkes\u00ebs lexuese. N\u00ebse nuk ka k\u00ebrkesa dhe ngarkes\u00eb \u2014 do t\u00eb \u00e7aktivizojm\u00eb replikat. Instanca e par\u00eb \u00ebsht\u00eb master, nd\u00ebrsa t\u00eb tjerat jan\u00eb replika.<\/p>\n
Kjo ide \u00ebsht\u00eb e realizuar n\u00eb baz\u00ebn e t\u00eb dh\u00ebnave t\u00eb quajtur Aurora Serverless AWS. Parimi \u00ebsht\u00eb i thjesht\u00eb: k\u00ebrkesat nga aplikacionet e jashtme pranohen nga flota e proxy. Duke par\u00eb rritjen e ngarkes\u00ebs, ai ndan burime llogarit\u00ebse nga instancat minimale t\u00eb p\u00ebrgatitura paraprakisht \u2014 lidhja realizohet sa m\u00eb shpejt t\u00eb jet\u00eb e mundur. \u00c7aktivizimi i instancave ndodh gjithashtu n\u00eb at\u00eb m\u00ebnyr\u00eb.<\/p>\n
Brenda Aurora ka konceptin e Nj\u00ebsis\u00eb s\u00eb Kapacitetit Aurora, ACU. Kjo \u00ebsht\u00eb (n\u00eb nj\u00eb kuptim) \u2014 nj\u00eb instanc\u00eb (server). \u00c7do ACU konkret mund t\u00eb jet\u00eb master ose slave. \u00c7do Nj\u00ebsi Kapaciteti ka memorien e tij operuese, procesorin dhe diskun minimal. P\u00ebr rrjedhoj\u00eb, nj\u00eb master, t\u00eb tjerat jan\u00eb replika vet\u00ebm p\u00ebr lexim.<\/p>\n
Numri i k\u00ebtyre Nj\u00ebsive t\u00eb Kapacitetit Aurora n\u00eb pun\u00eb \u00ebsht\u00eb nj\u00eb parametr i konfiguruar. Numri minimal mund t\u00eb jet\u00eb nj\u00eb ose zero (n\u00eb k\u00ebt\u00eb rast, baza nuk punon n\u00ebse nuk ka k\u00ebrkesa).<\/p>\n
Baza Aurora Serverless mund t\u00eb skaloj\u00eb ngarkes\u00ebn lexuese. Por n\u00eb dokumentacion kjo nuk thuhet qart\u00eb. Mund t\u00eb krijohet iluzioni se ata mund t\u00eb aktivizojn\u00eb multi-master. Nuk ka asnj\u00eb magji n\u00eb k\u00ebt\u00eb. <\/p>\n
Kjo baz\u00eb \u00ebsht\u00eb mjaft e p\u00ebrshtatshme p\u00ebr t\u00eb mos shpenzuar shum\u00eb para p\u00ebr sisteme me akses t\u00eb paparashikuar. P\u00ebr shembull, kur krijojm\u00eb MVP ose faqe marketingu, zakonisht nuk presim ngarkes\u00eb t\u00eb q\u00ebndrueshme. Prandaj, n\u00eb munges\u00eb akses, ne nuk paguajm\u00eb p\u00ebr instancat. Kur papritur ndodh ngarkesa, p\u00ebr shembull, pas nj\u00eb konference ose fushate reklamimi, turma e njer\u00ebzve hyn n\u00eb faqen dhe ngarkesa rritet ndjesh\u00ebm, Aurora Serverless automatikisht e pranon k\u00ebt\u00eb ngarkes\u00eb dhe lidhet shpejt me burimet e nevojshme (ACU). M\u00eb pas, konferenca kalon, t\u00eb gjith\u00eb e harrojn\u00eb prototipin, server\u00ebt (ACU) fik\u00ebn, dhe shpenzimet bien n\u00eb zero \u2014 shum\u00eb e p\u00ebrshtatshme.<\/p>\n
Ky zgjidhje nuk \u00ebsht\u00eb e p\u00ebrshtatshme p\u00ebr ngarkesa t\u00eb larta dhe t\u00eb q\u00ebndrueshme, sepse nuk din t\u00eb shkall\u00ebzoj\u00eb ngarkes\u00ebn e shkrimit. T\u00eb gjitha k\u00ebto lidhje dhe \u00e7 lidhje burimesh ndodhin n\u00eb momentin e ashtuquajtur 'scale point' \u2014 momenti kur databaza nuk mbahet nga nj\u00eb transaksion, nuk mbahen tabela t\u00eb p\u00ebrkohshme. P\u00ebr shembull, gjat\u00eb nj\u00eb jave scale point mund t\u00eb mos ndodhi, dhe databaza punon me t\u00eb nj\u00ebjtat burime dhe thjesht nuk mund t\u00eb vazhdoj\u00eb as t\u00eb zgjeroh\u00ebt, as t\u00eb t\u00ebrheq\u00eb. <\/p>\n
Nuk ka magji \u2014 ky \u00ebsht\u00eb nj\u00eb PostgreSQL i zakonsh\u00ebm. Por procesi i shtimit t\u00eb makinave dhe \u00e7aktivizimi \u00ebsht\u00eb pjes\u00ebrisht i automatizuar.<\/p>\n
Serverless me dizajn<\/h2>\n \nAurora Serverless \u00ebsht\u00eb nj\u00eb baz\u00eb e vjet\u00ebr, e ripar\u00eb p\u00ebr re, p\u00ebr t\u00eb p\u00ebrdorur avantazhet e ve\u00e7anta Serverless. Tani do t\u00eb flas p\u00ebr baz\u00ebn q\u00eb nga fillimi \u00ebsht\u00eb shkruar p\u00ebr re, p\u00ebr qasjen serverless \u2014 Serverless-by-design. Ajo \u00ebsht\u00eb zhvilluar menj\u00ebher\u00eb pa supozimin se do t\u00eb funksionoj\u00eb n\u00eb server\u00eb fizik\u00eb.<\/p>\n
Kjo baz\u00eb quhet Snowflake. Ajo ka tre blloqe kryesore.<\/p>\n
Blloku i dyt\u00eb \u2014 \u00ebsht\u00eb nj\u00eb shum\u00ebllojshm\u00ebri klasteresh virtuale p\u00ebr llogaritje (n\u00eb ilustrim \u2014 nj\u00eb grup rrethesh blu).<\/p>\n
Blloku i tret\u00eb \u2014 \u00ebsht\u00eb sistemi i ruajtjes s\u00eb t\u00eb dh\u00ebnave mbi baz\u00ebn S3. S3 \u00ebsht\u00eb nj\u00eb magazin\u00eb objektesh pa kufi n\u00eb AWS, di\u00e7ka si nj\u00eb Dropbox pa kufij p\u00ebr biznesin.<\/p>\n
Le t\u00eb shohim se si funksionon Snowflake, duke supozuar nj\u00eb fillim t\u00eb ftoht\u00eb. Pra, baza ekziston, t\u00eb dh\u00ebnat jan\u00eb ngarkuar n\u00eb t\u00eb, nuk ka k\u00ebrkesash aktive. K\u00ebshtu q\u00eb, n\u00ebse nuk ka k\u00ebrkesa p\u00ebr baz\u00ebn, ne kemi ngritur nj\u00eb sh\u00ebrbim t\u00eb shpejt\u00eb Metadata n\u00eb memorie (blloku i par\u00eb). Dhe kemi nj\u00eb depo S3, ku ndodhen t\u00eb dh\u00ebnat e tabelave, t\u00eb ndara n\u00eb ato q\u00eb quhen mikroparti. P\u00ebr thjesht\u00ebsi: n\u00ebse tabela p\u00ebrmban transaksione, mikroparti jan\u00eb dit\u00ebt e transaksioneve. \u00c7do dit\u00eb \u00ebsht\u00eb nj\u00eb mikroparti e ve\u00e7ant\u00eb, nj\u00eb skedar i ve\u00e7ant\u00eb. Dhe kur baza punon n\u00eb k\u00ebt\u00eb m\u00ebnyr\u00eb, ju paguani vet\u00ebm p\u00ebr hap\u00ebsir\u00ebn q\u00eb z\u00ebn\u00eb t\u00eb dh\u00ebnat. P\u00ebr m\u00eb tep\u00ebr, tarifa p\u00ebr hap\u00ebsir\u00ebn \u00ebsht\u00eb shum\u00eb e ul\u00ebt (sidomos duke marr\u00eb parasysh kompresimin e konsideruesh\u00ebm). Sh\u00ebrbimi i metadatas gjithashtu funksionon vazhdimisht, por p\u00ebr optimizimin e k\u00ebrkesave nuk nevojiten shum\u00eb burime, dhe sh\u00ebrbimi mund t\u00eb konsiderohet zhvillimsa-burim t\u00eb kushtit t\u00eb lir\u00eb. <\/p>\n
Tani imagjinoni q\u00eb nj\u00eb p\u00ebrdorues ka ardhur n\u00eb baz\u00ebn ton\u00eb dhe ka hedhur nj\u00eb k\u00ebrkes\u00eb SQL. K\u00ebrkesa SQL menj\u00ebher\u00eb d\u00ebrgohet p\u00ebr p\u00ebrpunim n\u00eb sh\u00ebrbimin Metadata. K\u00ebshtu, pasi merr k\u00ebrkes\u00ebn, ky sh\u00ebrbim analizon k\u00ebrkes\u00ebn, t\u00eb dh\u00ebnat e disponueshme, dhe autorizimet e p\u00ebrdoruesit dhe, n\u00ebse gjith\u00e7ka shkon mir\u00eb, p\u00ebrgatit nj\u00eb plan p\u00ebr p\u00ebrpunimin e k\u00ebrkes\u00ebs.<\/p>\n
M\u00eb pas, sh\u00ebrbimi fillon aktivizimin e klasterit t\u00eb llogaritjeve. Klasteri i llogaritjeve \u00ebsht\u00eb nj\u00eb klaster serverash q\u00eb kryejn\u00eb llogaritje. Kjo do t\u00eb thot\u00eb se mund t\u00eb p\u00ebrmbaj\u00eb 1 server, 2 servera, 4, 8, 16, 32 \u2014 sa t\u00eb doni. Ju d\u00ebrgoni nj\u00eb k\u00ebrkes\u00eb dhe menj\u00ebher\u00eb fillohet aktivizimi i k\u00ebtij klasteri. Realisht, kjo zgjat disa sekonda.<\/p>\n
\u00c7do k\u00ebrkes\u00eb SQL mund t\u00eb jo vet\u00ebm t\u00eb lidh\u00eb agregat\u00ebt nga t\u00eb dh\u00ebnat e ngarkuara m\u00eb par\u00eb, por gjithashtu t\u00eb ngarkoj\u00eb\/formoj\u00eb t\u00eb dh\u00ebna t\u00eb reja n\u00eb baz\u00eb. Kjo do t\u00eb thot\u00eb se kjo mund t\u00eb jet\u00eb nj\u00eb k\u00ebrkes\u00eb q\u00eb, p\u00ebr shembull, kryen inserimin e sh\u00ebnimeve t\u00eb reja n\u00eb nj\u00eb tabel\u00eb tjet\u00ebr, e cila \u00e7on n\u00eb krijimin e nj\u00eb ndarje t\u00eb re n\u00eb klas\u00ebn kompjuterike, q\u00eb, nga ana e saj, ruhet automatikisht n\u00eb depozitat e vetme S3.<\/p>\n
Skenari i p\u00ebrshkruar m\u00eb sip\u00ebr, nga ardhja e p\u00ebrdoruesit deri te ngritja e klasht\u00ebr\u00ebve, ngarkimi i t\u00eb dh\u00ebnave, ekzekutimi i k\u00ebrkesave, marrja e rezultateve, k\u00ebt\u00eb e paguani sipas tarif\u00ebs p\u00ebr minutat e p\u00ebrdorimit t\u00eb klasht\u00ebr\u00ebs virtuale t\u00eb llogaritur, warehouse virtual. Tarifa variolon n\u00eb var\u00ebsi t\u00eb zon\u00ebs AWS dhe madh\u00ebsis\u00eb s\u00eb klasht\u00ebr\u00ebs, por, n\u00eb mesatare, \u00ebsht\u00eb disa dollar\u00eb n\u00eb or\u00eb. Nj\u00eb klasht\u00ebr me kat\u00ebr makina \u00ebsht\u00eb dyfish m\u00eb e shtrenjt\u00eb se nj\u00eb me dy makina, nd\u00ebrsa nj\u00eb me tet\u00eb makina \u00ebsht\u00eb dyfish m\u00eb e shtrenjt\u00eb se ajo me kat\u00ebr. Disponohen opsione nga 16, 32 makina, n\u00eb var\u00ebsi t\u00eb kompleksitetit t\u00eb k\u00ebrkesave. Por paguani vet\u00ebm p\u00ebr ato minuta kur klasht\u00ebria \u00ebsht\u00eb n\u00eb pun\u00eb, sepse kur s'ka k\u00ebrkesa, ju si\u00e7 b\u00ebni heqjen dor\u00eb, dhe pas 5-10 minutash pritjeje (paramet\u00ebr i p\u00ebrshtatsh\u00ebm) ai automatikisht do t\u00eb ndalet, do t\u00eb liroj\u00eb burimet dhe do t\u00eb b\u00ebhet falas.<\/p>\n
\u00cbsht\u00eb krejt\u00ebsisht e mundur skenari, ku ju d\u00ebrgoni nj\u00eb k\u00ebrkes\u00eb, klasht\u00ebria shfaqet, po t\u00eb themi, p\u00ebr nj\u00eb minut\u00eb, nj\u00eb minut\u00eb tjet\u00ebr ajo llogarit, m\u00eb pas pes\u00eb minuta p\u00ebr fikje, dhe n\u00eb fund paguani vet\u00ebm p\u00ebr shtat\u00eb minuta t\u00eb pun\u00ebs s\u00eb k\u00ebsaj klashtre, e jo p\u00ebr muaj e vite.<\/p>\n
Skenari i par\u00eb p\u00ebrshkroi p\u00ebrdorimin e Snowflake n\u00eb nj\u00eb variant p\u00ebr nj\u00eb p\u00ebrdorues. Tani le t\u00eb imagjinojm\u00eb se ka shum\u00eb p\u00ebrdorues, q\u00eb \u00ebsht\u00eb m\u00eb af\u00ebr nj\u00eb skenari real.<\/p>\n
Le t\u00eb supozojm\u00eb se kemi shum\u00eb analist\u00eb dhe raporte Tableau, t\u00eb cil\u00ebt vazhdimisht bombardojn\u00eb baz\u00ebn ton\u00eb me nj\u00eb sasi t\u00eb madhe k\u00ebrkesh analitike SQL t\u00eb thjeshta.<\/p>\n
P\u00ebrve\u00e7 k\u00ebsaj, le t\u00eb supozojm\u00eb se kemi Data Scientists t\u00eb zellsh\u00ebm, q\u00eb p\u00ebrpiqen t\u00eb b\u00ebjn\u00eb gj\u00ebra monstruoze me t\u00eb dh\u00ebnat, t\u00eb operojn\u00eb me dhjet\u00ebra Terabajt, analizohet miliarda dhe triliona rreshta t\u00eb dh\u00ebnash. <\/p>\n
P\u00ebr llojet e ngarkesave t\u00eb p\u00ebrshkruara m\u00eb sip\u00ebr, Snowflake lejon ngritjen e disa klaster\u00ebve t\u00eb llogaritjes t\u00eb pavarur me fuqi t\u00eb ndryshme. Dhe k\u00ebta klaster\u00eb llogarit\u00ebs punojn\u00eb n\u00eb m\u00ebnyr\u00eb t\u00eb pavarur, por me t\u00eb dh\u00ebna t\u00eb p\u00ebrbashk\u00ebta t\u00eb harmonizuara.<\/p>\n
P\u00ebr nj\u00eb num\u00ebr t\u00eb madh k\u00ebrkesash t\u00eb lehta, mund t\u00eb krijoni 2-3 klastera t\u00eb vegj\u00ebl, secili me p\u00ebrmasat, p\u00ebr ngarkesa t\u00eb p\u00ebraf\u00ebrta, nga 2 makina. Ky funksionim \u00ebsht\u00eb i realizuesh\u00ebm, p\u00ebrfshir\u00eb edhe p\u00ebrmes konfigurimeve automatik. K\u00ebshtu, thoni: \"Snowflake, ngrit nj\u00eb klaster t\u00eb vog\u00ebl. N\u00ebse ngarkesa mbi t\u00eb kalon nj\u00eb paramet\u00ebr t\u00eb caktuar, ngrit nj\u00eb t\u00eb dyt\u00eb, t\u00eb tret\u00eb. Kur ngarkesa fillon t\u00eb bie, ndalo t\u00eb tep\u00ebrta.\" K\u00ebshtu, pavar\u00ebsisht n\u00ebse sa analist\u00eb vijn\u00eb dhe fillojn\u00eb t\u00eb shqyrtojn\u00eb raportet, t\u00eb gjith\u00eb kan\u00eb mjaft burime.<\/p>\n
N\u00eb t\u00eb nj\u00ebjt\u00ebn koh\u00eb, n\u00ebse analist\u00ebt flen\u00eb dhe askush nuk po sheh raportet, klasterat mund t\u00eb fik\u00ebn plot\u00ebsisht, dhe k\u00ebshtu ndaloni pages\u00ebn p\u00ebr to.<\/p>\n
P\u00ebr k\u00ebrkesa t\u00eb r\u00ebnda (nga Data Scientists), mund t\u00eb krijoni nj\u00eb klaster shum\u00eb t\u00eb madh me p\u00ebrmasat e p\u00ebraf\u00ebrta 32 makina. Ky klaster gjithashtu do t\u00eb paguhet vet\u00ebm p\u00ebr ato minuta dhe or\u00eb kur aty funksionon k\u00ebrkesa juaj gjigande.<\/p>\n
Mund\u00ebsia e p\u00ebrshkruar m\u00eb sip\u00ebr lejon ndarjen midis klasterave jo vet\u00ebm p\u00ebr 2, por edhe p\u00ebr m\u00eb shum\u00eb lloje ngarkesash (ETL, monitorim, materializimi i raporteve,\u2026).<\/p>\n
Le t\u00eb b\u00ebjm\u00eb nj\u00eb p\u00ebrmbledhje t\u00eb Snowflake. Baza kombinon nj\u00eb ide t\u00eb bukur me nj\u00eb realizim funksional. N\u00eb ManyChat ne p\u00ebrdorim Snowflake p\u00ebr analizimin e t\u00eb gjitha t\u00eb dh\u00ebnave t\u00eb disponueshme. Ne kemi nga 5 deri n\u00eb 9 klaster\u00eb t\u00eb ndrysh\u00ebm n\u00eb vend t\u00eb tre, si\u00e7 \u00ebsht\u00eb n\u00eb shembull. Kemi klaster\u00eb me 16 makina, me 2 makina, dhe madje edhe super-t\u00eb vogla me 1 makin\u00eb p\u00ebr disa detyra. Ata shp\u00ebrndajn\u00eb ngarkes\u00ebn me sukses dhe na lejojn\u00eb t\u00eb kursejm\u00eb shum\u00eb.<\/p>\n
Baza e Snowflake e shkall\u00ebzon me sukses ngarkes\u00ebn lexuese dhe shkruese. Ky \u00ebsht\u00eb nj\u00eb ndryshim i madh dhe nj\u00eb p\u00ebrparim i madh n\u00eb krahasim me \"Auror\u00ebn\", e cila p\u00ebrballonte vet\u00ebm ngarkes\u00ebn lexuese. Snowflake lejon t\u00eb shkall\u00ebzohet edhe ngarkesa shkruese me k\u00ebto klaster\u00eb llogarit\u00ebs. Pra, si\u00e7 e p\u00ebrmenda, ne n\u00eb ManyChat p\u00ebrdorim disa klaster\u00eb; klaster\u00ebt e vegj\u00ebl dhe super-t\u00eb vegj\u00ebl p\u00ebrdoren kryesisht p\u00ebr ETL, p\u00ebr ngarkimin e t\u00eb dh\u00ebnave. Nd\u00ebrsa analist\u00ebt punojn\u00eb n\u00eb klaster\u00ebt mesatar\u00eb, t\u00eb cil\u00ebt nuk preken nga ngarkesa ETL, prandaj punojn\u00eb shum\u00eb shpejt. <\/p>\n
K\u00ebshtu, baza \u00ebsht\u00eb e p\u00ebrshtatur mir\u00eb p\u00ebr detyrat OLAP. Megjithat\u00eb, p\u00ebr ngarkesat OLTP, ajo ende nuk \u00ebsht\u00eb e aplikueshme. S\u00eb pari, kjo baz\u00eb \u00ebsht\u00eb kolonale, me t\u00eb gjitha pasojat q\u00eb vijn\u00eb nga kjo. S\u00eb dyti, vet\u00eb qasja, kur p\u00ebr \u00e7do k\u00ebrkes\u00eb ngat\u00ebrroni nj\u00eb kluster llogarit\u00ebs dhe e mbushni at\u00eb me t\u00eb dh\u00ebna, fatkeq\u00ebsisht, p\u00ebr ngarkesat OLTP nuk \u00ebsht\u00eb ende mjaft e shpejt\u00eb. Disa sekonda pritjeje p\u00ebr detyrat OLAP jan\u00eb t\u00eb pranuara, ndryshe p\u00ebr detyrat OLTP \u2014 \u00ebsht\u00eb e papranueshme, do t\u00eb ishte m\u00eb mir\u00eb 100 ms, dhe akoma m\u00eb mir\u00eb \u2014 10 ms.<\/p>\n
P\u00ebrfundimi<\/h2>\n \nBaza e t\u00eb dh\u00ebnave pa server \u00ebsht\u00eb e mundur p\u00ebrmes ndarjes s\u00eb baz\u00ebs n\u00eb pjes\u00eb Stateless dhe Stateful. Ju, sigurisht, keni v\u00ebn\u00eb re se n\u00eb t\u00eb gjitha shembujt e paraqitur, pjesa Stateful \u00ebsht\u00eb, n\u00eb nj\u00ebfar\u00eb m\u00ebnyre, ruajtja e mikro-particioneve n\u00eb S3, nd\u00ebrsa Stateful \u00ebsht\u00eb optimizuesi, punimi me metadatat, p\u00ebrpunimi i \u00e7\u00ebshtjeve t\u00eb siguris\u00eb, t\u00eb cilat mund t\u00eb ngrihen si sh\u00ebrbime t\u00eb lehta Stateless t\u00eb pavarura.<\/p>\n
Kryerja e k\u00ebrkesave SQL gjithashtu mund t\u00eb perceptohet si sh\u00ebrbime me nj\u00eb gjendje t\u00eb leht\u00eb, t\u00eb cilat mund t\u00eb ngrihen n\u00eb modalitetin pa server, si klasteret llogaritus Snowflake, p\u00ebr t\u00eb shkarkuar vet\u00ebm t\u00eb dh\u00ebnat q\u00eb nevojiten, p\u00ebr t\u00eb kryer k\u00ebrkes\u00ebn dhe \"t\u00eb shuajn\u00eb\".<\/p>\n
Baza serverless t\u00eb nivelit t\u00eb prodhimit jan\u00eb tani t\u00eb disponueshme p\u00ebr p\u00ebrdorim dhe funksionojn\u00eb. K\u00ebto baza serverless jan\u00eb tashm\u00eb t\u00eb gatshme p\u00ebr t'u p\u00ebrballur me detyra OLAP. Megjithat\u00eb, p\u00ebr detyrat OLTP ato p\u00ebrdoren\u2026 me disa nuanca, pasi ka kufizime. Nga nj\u00ebra an\u00eb, kjo \u00ebsht\u00eb nj\u00eb mang\u00ebsi. Por, nga ana tjet\u00ebr, kjo \u00ebsht\u00eb nj\u00eb mund\u00ebsi. Ndoshta ndonj\u00eb nga lexuesit do t\u00eb gjej\u00eb nj\u00eb m\u00ebnyr\u00eb p\u00ebr ta b\u00ebr\u00eb baz\u00ebn OLTP plot\u00ebsisht serverless, pa kufizime Aurora.<\/p>\n
Shpresoj se ka qen\u00eb interesante p\u00ebr ju. P\u00ebr nj\u00eb t\u00eb ardhme serverless \ud83d\ude42habr.com<\/a> <\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"\u0412\u0441\u0435\u043c \u043f\u0440\u0438\u0432\u0435\u0442! \u041c\u0435\u043d\u044f \u0437\u043e\u0432\u0443\u0442 \u0413\u043e\u043b\u043e\u0432 \u041d\u0438\u043a\u043e\u043b\u0430\u0439. \u0420\u0430\u043d\u044c\u0448\u0435 \u044f \u0440\u0430\u0431\u043e\u0442\u0430\u043b \u0432 \u0410\u0432\u0438\u0442\u043e \u0438 \u0448\u0435\u0441\u0442\u044c \u043b\u0435\u0442 \u0440\u0443\u043a\u043e\u0432\u043e\u0434\u0438\u043b Data Platform, \u0442\u043e \u0435\u0441\u0442\u044c \u0437\u0430\u043d\u0438\u043c\u0430\u043b\u0441\u044f \u0432\u0441\u0435\u043c\u0438 \u0431\u0430\u0437\u0430\u043c\u0438: \u0430\u043d\u0430\u043b\u0438\u0442\u0438\u0447\u0435\u0441\u043a\u0438\u043c\u0438 (Vertica, ClickHouse), \u043f\u043e\u0442\u043e\u043a\u043e\u0432\u044b\u043c\u0438 \u0438 OLTP (Redis, Tarantool, VoltDB, MongoDB, PostgreSQL). \u0417\u0430 \u044d\u0442\u043e \u0432\u0440\u0435\u043c\u044f \u044f \u0440\u0430\u0437\u043e\u0431\u0440\u0430\u043b\u0441\u044f \u0441 \u0431\u043e\u043b\u044c\u0448\u0438\u043c \u043a\u043e\u043b\u0438\u0447\u0435\u0441\u0442\u0432\u043e\u043c \u0431\u0430\u0437 \u0434\u0430\u043d\u043d\u044b\u0445 \u2014 \u0441\u0430\u043c\u044b\u0445 \u0440\u0430\u0437\u043d\u044b\u0445 \u0438 \u043d\u0435\u043e\u0431\u044b\u0447\u043d\u044b\u0445, \u0438 \u0441 \u043d\u0435\u0441\u0442\u0430\u043d\u0434\u0430\u0440\u0442\u043d\u044b\u043c\u0438 \u043a\u0435\u0439\u0441\u0430\u043c\u0438 \u0438\u0445 \u0438\u0441\u043f\u043e\u043b\u044c\u0437\u043e\u0432\u0430\u043d\u0438\u044f. \u0421\u0435\u0439\u0447\u0430\u0441 […]<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"author":1,"featured_media":91636,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[688],"tags":[],"class_list":["post-91635","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-administrirovanie"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t