Zonke iinkampani ezisebenza Snowflake okanye BigQuery uyazi le ndingathanda: i-akhawunti yenyanga ifumaneka, i-40% ephakeme kunokwenzeka, kwaye akukho umntu angakwazi ukucacisa ngexabiso ngexabiso. Iinkompyutha zeData zithumela iiyure ezininzi angu-10 ngeveki ngokuthe ngamanzi ngaphandle kokwakha iipowduzi. Izixhobo zihlangana ukuhlaziywa, iinkqubo zihlangana ukusebenza, kunye nesisombululo ye-default, ukuxhaswa kwixesha elinye, ukuxhaswa kwakhona, kwaye oku kuthatha i-problem! Ido Arieli Noga, CEO kunye ne-Co-Founder Ukususela kwiminyaka engaphezu kwama-12, i-Snowflake iye yahlala kwiintsuku ezininzi zokusebenza kwiinkalo zeendaba ezininzi kwi-government, emva koko njenge-Head of Data kwi-Lightico apho uye uyakhawuleza i-Snowflake i-cost spiral. I-Yuki Kule nqakraza, sinxibelelana ngoko ke indlela yokufakelwa kwe-infrastructure ye-data ifumaneka, yintoni iinkqubo ze-AI-era zibonisa iibhanki ze-cloud, kunye neendlela ezininzi i-industry kufuneka ukutshintsha ukulawula iimfuno ze-data ukusuka kwakhona. Hi Ido, kakuhle kwi- "Behind the Startup" series yethu. Uyafuna kwi-government, i-DevOps, i-data architecture, kwaye ekugqibeleni i-Head of Data kwi-Lightico. Oku yindlela elide. Yintoni iingxaki zihlanganisa ukuqala i-Yuki? Ishan Pandey: I-Head of Data inxaxheba kwindawo yokuphindaphinda. I-team yam iye yenza i-60% yayo ixesha kwi-infrastructure - i-storage size, i-cost firefighting, i-tuning requests that shouldn't have needed tuning. Ido Arieli Noga: Ndingathanda ukuba ingxaki yenzelwe. Ndingathanda ukuba yenzelwe kwi-DevOps - iinkampani ezifana ne-Spot.io zibonise i-optimization ye-autonomous ye-cloud infrastructure kwaye iye yenzelwe. Yinto efanayo ayikho kwi-data platforms. Ngoko ke ndingathanda kwi-Yakir Daniel, i-co-founder ye-Spot.io, ndinika i-idea ezisetyenziswe kwi-data, kwaye uye kwenziwe ngexesha elide. Le nqakraza yandiswa kwimali yethu yokuqala kunye ne-validator enzima. Emva koko sikhumbule. Ngaba uqhagamshelane nge-400 iindidi zeendatha kwi-Snowflake, i-BigQuery kunye neDatabricks. Yintoni i-thread efanayo? Ishan Pandey: I-frustration efanayo, ngaphandle kwe-platform. Iingcali ziyazi ukuba zithunyelwe kakhulu. Zibonele. Kodwa ukulungiselela kufuneka i-auditing epheleleyo - i-magazine ye-magazine, i-workload ye-workload, kwaye akukho umntu unayo ububanzi we-bandwidth. Ido Arieli Noga: Iiplatforms ngokufanayo zinika iinketho ezincinane. Unako ukuguqulwa i-storage, kodwa oku yi-instrument emnyama. 
 
 Ukusebenza okwenziwe ngokwenene kwinqanaba le-query - zonke i-query zihlanganisa iimfuno ezahlukeneyo ze-computing, iimveliso ze-concurrency ezahlukeneyo, i-timing ezahlukeneyo. Unokufumaneka oku ngokuchofoza i-slider. Ufuna into efana ne-query ngamnye ngokufanelekileyo kwaye ibonelela i-infrastructure efanelekileyo ngexesha elifanelekileyo. Ukusebenza okwenziwe ngokwenene kwinqanaba le-query - zonke i-query zihlanganisa iimfuno ezahlukeneyo ze-computing, iimveliso ze-concurrency ezahlukeneyo, i-timing ezahlukeneyo. Unokufumaneka oku ngokuchofoza i-slider. Ufuna into efana ne-query ngamnye ngokufanelekileyo kwaye ibonelela i-infrastructure efanelekileyo ngexesha elifanelekileyo. I-200+ abaphakeli kule ndawo kubandakanya iibhodi ezininzi. I-gap yinto efanelekileyo - ngokuqhubekayo, kwinqanaba le-query, ngaphandle kokuhamba i-code ye-user. I-Storage Spread, Ukusetyenziswa kwe-Slot, Ukukhula kwe-Cluster - iiplatform ezahlukileyo, inkinga elifanayo. Yintoni kunokwenzeka? Ishan Pandey: I-provisioning iyafumaneka. I-right-sizeing iyafumaneka. Iqela leyo kufuneka ifumaneke i-use case entsha, i-spin-up, kwaye ifumaneke. Unomntu akhawunti ukusetyenziswa kwezinto ezininzi ngexesha elide - i-workload ifumaneka ngenyanga ngamnye nangona kunjalo. Ido Arieli Noga: Ukubona oku ngokugqithisileyo: kwi-Snowflake, i-40% yeengxaki ze-medium warehouse ingqalelo kwi-extra-small. kwi-BigQuery, i-slot reservations yi-routine ye-3-4x i-reel demand. Iingcali akufutshane - akukho ingxaki eyenziwe ngempumelelo ngempumelelo. Ngaba uqhagamshelane nge-time-to-value fast. Yintoni ukuqhagamshelwano olufanelekileyo, kwaye ngoko kufuneka ifumaneke? Ishan Pandey: Thina ndingathanda phakathi kwe-application kunye ne-data platform kwinqanaba le-connection. Thina ndingathanda zonke iingxaki ngexesha elifanelekileyo - ubungakanani, ukuxhaswa, ukuxhaswa, isakhiwo se-history kwaye uqhagamshelane kwi-computing ye-size elifanelekileyo. Akukho ukuguqulwa kwe-code njengoko hhayi kuxhaswa iingxaki. Ido Arieli Noga: Isantya sokuphendula kuza kuxhomekeke kwinqanaba elide esekelwe kwiimeko ezininzi. Iinkcukacha ezininzi ziyafumaneka kakhulu ukususela ngosuku lokuqala. 
 
 I-Angel Studios ifumaneka kwi-54 imizuzu kunye ne-connection string swap kunye ne-60% ukunciphisa ixabiso. I-Tenable ibandakanya i-33% kunye ne-25% yexesha le-engineering kwiiveki ezimbini. I-Angel Studios ifumaneka kwi-54 imizuzu kunye ne-connection string swap kunye ne-60% ukunciphisa ixabiso. I-Tenable ibandakanya i-33% kunye ne-25% yexesha le-engineering kwiiveki ezimbini. Kwimeko apho akufanelekileyo: iinqwelo ezininzi ezidlulileyo kwi-manual optimization, iingxaki zokusebenza ezininzi ezincinane ngaphandle kweemveliso ezincinane, okanye iinkalo apho iindleko ezininzi zihlanganisa okanye ukuhambisa idatha kunokuba yi-computing. I-Priceing yakho yinkqubo ye-savings yayo - akukho i-savings, akukho i-charge. Yinto engapheliyo kwi-Enterprise SaaS. Yintoni, kwaye njani i-CFO? Ishan Pandey: Ukusuka kwimvumelwano. Ukuba iimveliso isebenza, ukuxhaswa ukhuseleko kuyimfuneko. Ukuba akufanelekileyo kwimeko esifunyenweyo, akufuneka ukuxhaswa. Ido Arieli Noga: Kwimeko yeCFO, inkxaso lula. Siza kuqhuba iintsuku ze-30 ze-pilot ngexabiso. Ukuba awukwazi ukuvelisa, akukho i-akhawunti. Oku ukunciphisa ngokupheleleyo i-risk ye-evaluation. I-part enzima kakhulu yi-procurement - ezinye iinkampani ze-finance ayikho ibhokisi eluhlaza ye-SaaS esekelwe kwimiphumo. Kodwa xa zibonise, kubandakanya i-deal elula ukuvumelela. I-AI i-experimentation yenza inguqulelo entsha ye-computing waste problem. Yintoni ukuguqulwa izinto? Ishan Pandey: Iintlobo zokusebenza ze-AI ziquka kunye nexesha elidlulileyo. I-team isebenze i-experiment, ifumaneka okanye ifumaneka, kwaye i-computing isebenza. Ukubandakanya oku kwiintlobo ezininzi kunye ne-waste compounds ngokukhawuleza. Ido Arieli Noga: I-Workload-aware layer yandisa i-calculus. Iingoma ziyafumaneka i-workloads ngaphandle kwe-cost-anxiety ebonakalayo ngenxa yokufunda i-infrastructure. Oku kubonakala ingxaki engapheliyo, kodwa ivimbela i-point ye-friction ye-real - uye uyakwazi ukufumana "umayeza ukuqhuba oku?" kunye nokufunda ukufumana i- "Ukufundisa njani oku?" Kwakhona inikeza umlawuli weedatha ukubonakalisa apho intengiso ye-AI ithatha - ngexesha le-team, i-workload, i-outcome. Oku kwenza i-governance layer, kwaye akuyona i-cost tool kuphela. Kuba umlawuli weData uqhagamshelane neengxaki ezininzi kunye neengcali abalandeli kwi-infrared work - iintlobo ezintathu zihlanganisa ngeveki, kwi-plattform eyona. Ishan Pandey: Okokuqala, ukusetyenziswa kwe-computing yi-time window. Izixhobo ezininzi ziyafumaneka kwi-off-peak hours. Fast to find, fast to fix. Ido Arieli Noga: Okokuqala, i-resource level vs. i-request demand. Iintsuku ze-30 ze-history ye-query, eziqhelekileyo yi-calculation level. Ukuba i-30% ezininzi ze-query zithunyelwe phantsi kwe-capacity yayo, unayo ingxaki ye-size. Okwangoku, iimveliso ever-on ngaphandle kokuvelisa ukusebenza. Iimveliso, iinkcukacha, okanye iinkcukacha zokuzonwabisa ukuba awukwazi ukuxhaswa ngokupheleleyo - ngokuvamile zihlanganiswa ngempumelelo okanye zithintshwe ngexesha lokugqibela. Iimveliso ephakamileyo, zithintshwa ngokukhawuleza. Ngaba utshintshe ngaphezu kwe-Snowflake ukuya kwi-BigQuery kunye ne-Databricks. Kwaye le ngqela elandelayo? Ishan Pandey: I-Vision yi-optimization layer eyodwa kwiinkqubo zeendaba ezininzi ze-Data Platforms - akuyona i-Snowflake tool, akuyona i-cost tool. I-Resource management, i-observability, i-governance, kwi-platform ye-Snowflake. Ido Arieli Noga: I-analogue ye-auto-escalation ithatha. Ngoku, ukuqhuba i-data platform ngaphandle kwe-optimization layer kunceda ukhuseleko kwiingcali ukwenza into efunyenweyo. Emva kwiminyaka emininzi oku kuya kubona efana ne-manual server provisioning ngexesha elandelayo - akukho ingxaki, nje engabonakali. Ukuguqulwa kwakhona. I-Infrastructure yeData yaye ngoku i-infrastructure ye-business core. I-tooling eyenziwe ngoko kufuneka ifumaneke oku. Ngaba ushiye ukuthetha kwaye ushiye iinkcukacha!

This story contains new, firsthand information uncovered by the writer.

Hot off the press! This story contains factual information about a recent event.

Between Two Computer Monitors: This story includes an interview between the writer and guest/interviewee. 

Winner of Noonies 2020 for the award Hacker Noon Contributor of the Year - DATA.

2020 - Hacker Noon Contributor of the Year in DATA

Read My Stories

Le audio iveliswe ngolwimi lokuqala lwebali!

Yuki CEO kwi-Data Cost Optimization: Yintoni i-Snowflake yakho kunye ne-BigQuery Bill i-40% kunye nokukhula ngenxa ye-AI

About Author

IZIMVO

ZIJONGE IIMPAWU

ELI NQAKU LINIKEZELWE KU

Related Stories

Light-Mode

Classic

Newspaper

Minty

Dark-Mode

Neon Noir

Minty

HN StartUps