Kwi-world ye-digital-first namhlanje, ukulungiselela ukufikelela kwedatha ngokushesha kunokuba kuyimfuneko - kuyimfuneko. Ukuba usebenza isicelo ebalulekileyo, ukunikezela iinkcukacha ze-multimedia kwinqanaba yehlabathi yehlabathi, okanye nje ukusika iifayile ze-private, ukufikelela kwe-cloud storage yakho ikhawuleza ngqo yonke into, ukusuka kwimveliso ukuya kwi-confidence. Kodwa yintoni i- "i-reliable" ikakhulukazi? Kwimveliso ye-cloud, ukucaciswa kwinqanaba kwi-uptime percentages — 99.9%, 99.99%, okanye ngisho Nangona kunjalo, phantsi kwii-SLAs (i-Service Level Agreements) i-reality emangalisayo: - Umthamo ukufikelela idatha yakho ngexesha elinye, kwindawo eyodwa, ngaphandle kwemibuzo okanye iingxaki ezidlulileyo-ukuzithintela. Kwakhona i-centralized clouds emangalisayo into efanelekileyo into efanelekileyo: isixeko. five nines continuous performance I-Reliability ayikho into efunyenwe - yinto esithile ukuba uyenze. Ukusuka kwiindawo ezininzi ukuya kwiindlela zeenethiwekhi ezincinciphekileyo, siye sikholelwa ukuba i-infrastructure ye-centralized, nangona i-fortified, ayikwazi ukufumana izibambiso zayo zayo zayo zokusebenza. Xa zonke iindlela zihamba ngeenkcukacha zeenkcukacha ze-data centers, ingxaki esisodwa - i-fire, i-routing problem, i-internal error - ingaba iimpembelelo ze-riple kwiinkalo ezininzi. Ukusuka kwiindawo ezininzi ukuya kwiindlela zeenethiwekhi ezincinciphekileyo, siye sikholelwa ukuba i-infrastructure ye-centralized, nangona i-fortified, ayikwazi ukufumana izibambiso zayo zayo zayo zokusebenza. Xa zonke iindlela zihamba ngeenkcukacha zeenkcukacha ze-data centers, ingxaki esisodwa - i-fire, i-routing problem, i-internal error - ingaba iimpembelelo ze-riple kwiinkalo ezininzi. Intshayelelo le-blog ibonelela ukuba i-architecture ye-Sia ye-decentralized ibonwe ngokufanelekileyo ukuxhaswa kwakhona. Ngaphezulu kwe-Sia kune-platform ye-storage, i-Sia yenzelwe ukuqinisekisa ukuba Kwizigaba ezidlulileyo, siya kubandakanya le design kunye nemodeli zokuzonwabisa, ukunciphisa iingxaki zehlabathi, kwaye ibonise ukuba i-decentralization ayikho kuphela kunzima kakhulu - kunzima kakhulu. Ukusebenza nge-resilience Ngenxa yokuba kwihlabathi ye-cloud storage, ubonakalelo ayikho into olufunayo - yinto eyenzayo. Izifundo ezincinane ze-centralized clouds Ukubonisa yonke imibuzo yayo ye-5 ye-9 ye-uptime, i-cloud storage platform ye-traditional ibonise ngokuqinisekileyo ukuba i-infrastructure ye-centralized inokufanelekileyo xa ifumaneka kwi-environmental extremes, i-human error, okanye i-internal misconfigurations. Nangona i-redundancy ebizwa yi-hyperscaleers ezifana ne-AWS, i-Google Cloud, kunye ne-Microsoft Azure, izifundo zeemeko ze-real-world ibonise umzekelo olungapheliyo - enye apho iimilion yeamakhasimende inokuzalwa ukufikelela kwimeko, kwaye iimveliso ezininzi zithunywe kwi-stop-down ngenxa yeenkcukacha esisodwa. Umzekelo we-cloud ye-fragility kunokwenzeka kwizinto ze-datacenter fires - iingxaki ezinikezele ngexesha lokugqibela kwiinkonzo ze-cloud. Kwi-Agasti ye-2022, i-explosion ye-electricity kwi-Council Bluffs data center ye-Google iye yenza abasebenzi ezintathu kwaye zithintela iinkonzo ze-core ezifana ne-Search kunye ne-Maps. I-event, i-arc flash eyenziwe ngexesha lokusebenza kwe-substation, isebenza njengomngcele ukuba amaxabiso ze-industry angakwazi ukufumana iingxaki ezinxulumene ne-physical infrastructure. Inyanga elandelayo eParis, i-multi-cluster failure kwi-europe-west9-a zone ye-Google Cloud yaqala kunye ne-water intrusion - ngokwemiphumo ye-colding system failure eyenza i-battery room kunye nokuvula i-fire. Le i-cascading failure ayikho kuphela kwi-Google's major cloud regions eYurophu, kodwa nangaphezulu kwe-90 iinkonzo ze-cloud ngexesha elide. Zifumaneka kwi-incidente ezidlulileyo ye-2021 OVHcloud e-Strasbourg, eyayithintela ngokupheleleyo i-datacenter ye-SBG2 kwaye zithintela ezinye ezininzi kwi-campus efanayo. Many customers had no disaster recovery plans in place, and entire websites were lost without backups. Ngomhla ngoJulayi 2022, i-record-breaking temperatures engaphezulu kwe-40 ° C (104 ° F) eLondon iye yenza iinkampani ze-Google kunye ne-Oracle data centers ngaphandle kwe-off-line ngenxa yeengxaki ze-cooling system. I-Google yaba kufuneka ifake iinkcukacha ze-cloud yokuvimbela iimpahla ze-hardware - iingcebiso elidlulileyo ukuba iimeko kuphela ingathintela iinkcukacha ze-service. Nangona kunjalo, akukho zonke iingxaki ziye ziye ziye ziye zithunyelwe kwi-catastrophe ye-physical – ezinye ziyi-catastrophe ye-digital ezidlulileyo. KwiFebruwari 2024, i-Google Cloud iye yenza ingxaki engaphezulu xa i-metadata store ye-regional iye yenza i-US-West1 kwisithunyelwe kwiiyure ezilixhenxe ezintathu. Ngokusho, ukuhlaziywa kwe-CrowdStrike ngoJulayi 2024 iye yenza iingxaki ezininzi ze-Microsoft Windows, okuholela kwiizigidi zolide zololiwe kunye neengxaki ezininzi zokuvelisa kwizimboni. Kwaye inethiwekhi ye-Content Delivery Network (CDN) efana ne-Fastly ibonelela kwi-2021, iye yenza iingxaki zehlabathi, ezivela kwi-Reddit, i-Spotify, kunye neengxaki zeendaba ezininzi kwisithuba ezininzi. I-cause? I-mono-culture ye-CDN yabasetyenziswa kwihlabathi. Ukusebenza okuqhubekayo nge-Design Nangona abaphakeli be-centralized ye-cloud zibonisa izixeko ezininzi ukhuseleko ukhuseleko, i-Sia ibandakanya inkcazelo ngokugqithisileyo ngokugqithiselwa kwimodeli ye-fortress. Ngaphandle kokubeka yonke into kwi-resilience ye-one-area okanye i-facility, i-Sia ibandakanya idatha yakho kwihlabathi, kwiintlobo ezininzi ze-nodes eziholise ngokuhambisana, usebenzisa i-matematics - akayi-marketing - ukhuseleko. Akukho kuphela i-infrastructure eyahlukileyo - kukho i-philosophy eyahlukileyo. Ukuvunyelwa okuvumela I-redundancy isetyenziswa ngokuqhelekileyo njengoko ukhuseleko yokhuseleko - indlela yokhuseleko kwi-failure. Kodwa kwi-Sia, kunokuba ngaphezulu. I-redundancy yenza ukusebenza okuqhubekayo. Ngokufanelekileyo, i-Sia ibandakanya zonke iifayile kwi-30 i-shards encrypted usebenzisa i-coding yokugqithisa. Kukho kuphela i-10 i-shards ezininzi ziquka ukuguqulwa ngokupheleleyo le ifayile. Oku kuthetha ukuba inethiwekhi unokuthintela kuphela iingxaki, kodwa ukusebenza okuguqithisileyo kwizithuthi ezivela ngamnye - konke ngelixa ukufumana ukufikelela okuzenzakalelayo. I-redundancy is a fallback – it is the foundation of continuous performance. Ngokuya, i-cloud ye-traditional ibonelela kwi-file replication epheleleyo kwiindawo ezincinane. Ukuba i-region eyona, ukufikelela kubhalwe okanye isakhelo - kwaye isitoreji eyongezelelweyo akufanele isantya olungcono. Iimodeli ye-Sia ifumaneka ngexesha elifanelekileyo. Iintlobo ze-Retrieval ziquka ngokuchasene ne-host availability kunye ne-network conditions - akukho ama-failovers, akukho ama- bottlenecks, akukho ama-stoptime windows. Nangona i-centralized cloud ingasetyenziswa kwe-coding yokugqithisa kwindawo ye-internal, zonke i-infrastructure ze-cloud ye-cloud ye-centralized ye-cloud ye-cloud ye-cloud ye-cloud ye-cloud ye-cloud ye-cloud ye-cloud ye-cloud ye-cloud ye-cloud. I-hosts ye-Sia, ngokufanayo, ifakwe ngokufanelekileyo - ngokuvamile kwabasebenzi ezahlukeneyo. Ukusebenzisa i-Sia yinto efana nokuhlanganisa idatha yakho kwi-30 i-cloud ezahlukeneyo ngokufanelekileyo. Akukho inkampani eyodwa ekulawuleni inkqubo, kwaye akukho indawo eyodwa yokukhangisa kunokufumana. Ukuvuthwa ngaphandle kwe-interruption Kwiimeko ezininzi ze-cloud, xa kukho iziphumo, ukusebenza kubaluleke. Nangona iinkqubo ze-failover ziyafumaneka, ukuchithwa kubandakanya isantya esincinane, ukufikelela kwimifanekiso, okanye ixesha lokugqithisa jikelele ngelixa i-infrastructure ingxaki ukufumana. I-Architecture ye-Sia isebenza ngokufanayo. Xa i-host enikezela iinkcukacha zakho kwi-offline-kuye ngenxa ye-fault, ukhuseleko, okanye ukucaciswa-iifayile zakho ziyafumaneka ngokupheleleyo. Akukho i-spinner yokuthumela, akukho i-sync lag, akukho i-alert. I-network ibhale i-shards eziluncedo kwi-hosts ezininzi, ukhethe ngokugqithisileyo iinketho ezininzi ezinikezwayo. Zonke oku kwenziwa phantsi komhlaba. Ngexesha elandelayo, kwi-background, i-renter software ikwazi ukuguqulwa ngokupheleleyo ngokufanelekileyo ngokufaka i-shards ezintsha kwi-hosts ezifanelekileyo. Le nqubo yokuguqulwa kwe-self-healing akufuneka nje ukusombulula kwiziphumo ezininzi - ibonelela ukuba imveliso ibekwe ngempumelelo. Sia akuyona nje ukususela ukususela - it isebenza ngokusebenzisa. Ngaphandle kokuphendula kwi-failure emva kokufumana, i-Sia ithatha i-churn njenge-comportment efunyenweyo - nto le-network yenzelwe ukuxhaswa ngempumelelo. It is a model of resilience that not only avoids disruptions - it actively shields users from even noticing them. Akukho point ye-failure I-centralized cloud platform ihambelana neengxaki ze-cascading ngenxa yokufikelela kwinkqubo ye-centralized. I-router efanelekileyo, i-deployment ye-software ye-defective, okanye i-power problem kwi-installation eyodwa, ingathintela kumazwe ngamazwe ngamazwe. I-Architecture ye-Sia ibonelela le mthengi ngokucacileyo. Akukho i-master node. Akukho i-central region. Akukho i-privileged authority enokufumana i-offline ye-system. Ngoku, idatha yakho yathunyelwa kwiintlobo ze-hosts ezininzi ehlabathini lonke - zonke zihlanganisa kuphela i-shards encrypted. Ukuba umnqweno omnye utshintshwe, isisombululo esebenza. Ukuba amashishini ezisixhenxe utshintshwe, ke utshintshwe. Akukho kufuneka "ukushintshwe" ngenxa yokuba akukho indlela yokuqala. No region. No master node. No bottleneck. Just unstoppable ukufikelela. Ukuphazamiseka kwe-centric dependency akuyona kuphela ukwandisa ukuphazamiseka kwe-fault tolerance - it Ngaba ungathanda ukuba i-region ifumaneke kwi-intanethi. Ngaba ungathanda i-gateway enezibophelele okanye inkonzo yokuguqulwa kwe-human administrator. Ungathanda idatha ukusuka kwindawo efanelekileyo kakhulu - ngokuqhelekileyo. prevents performance blackouts Ukucaciswa ngenxa yobugcisa, akukwazi kuphela Xa ushiye malunga ne "i-cloud reliability", sinokuthengiswa ngempumelelo - i-SLA ebizwa ngeengxaki zentengiso, iinkcukacha zokusebenza ze-uptime, kunye ne-brand reputation. Kodwa njengoko sisaziwa, kwakhona iinkonzo ze-cloud ezininzi ayikwazi ukufumana i-fragility ebonakalayo kunye ne-centralization. Ukuba i-fire, i-heat waves, okanye i-software missteps, i-cloud yobungcali yaye ngokuqinisekileyo iimveliso ezincinane ze-cascading ezisuka kwi-disruption yehlabathi. I-Sia isetyenzise indlela eyahlukileyo. Ngaphandle kokufumana ukuba i-infrastructure iya kukufumana kunye nokufumana umdla xa akufanele, i-Sia isetyenzise ukuba umdla inokufumaneka - kwaye ibekwe inkqubo evuselela ukusebenza nangona kunjalo. Oku kubalulekileyo. : akukho iinkonzo ezihlabathi, akukho iinkonzo ezihlabathi, akukho i-vendor lock-in. I-auto-repair, i-infrastructure ye-decentralized enikezela idatha yakho yokufumaneka ngenxa yokuba akukho umdlali omnye inokukwazi ukuvelisa. Ukusebenza okuqhubekayo Yinto engaphezulu kunyango lwezobugcisa. Yinto ukuguqulwa kwindlela yethu yokufunda umgangatho we-digital. Ngaphandle kokwakha iingcingo ezininzi kunye neengqongqo ezininzi, i-Sia ibandakanya iingcingo zayo. I-Sia ibandakanya ukhuseleko. Kwaye ngokwenza oku, i-Sia ibandakanya indlela ye-cloud storage enokwenzeka kwihlabathi apho ixesha le-stoppage ayikho kwakhona. Njengoko iinkampani ziyafumaneka izixeko ezininzi, izindleko ezininzi, kunye nezidingo zokuxhomekeka ezininzi, i-decentralization iye yaba ngaphezulu kakhulu - iyona elungileyo. Ukuba ubumnandi we-cloud ibekwe ukwakha iinkonzo ezibonakalayo Ngexesha elide, ngoko ke, ngoko ke ixesha siphinde ukucacisa malunga ne-trust kunye nokufunda ukucacisa malunga ne-certainty. Ukusebenza kuphela Nge-Sia, ukusebenza okuqhubekayo ayinxalenye imiqondiso. Kuyinto ukhuseleko. iinkcukacha Data Center Knowledge. (2022, 9 EyeThupha). Data Center umdla — Google unemibuzo ‘electric incident’, 3 abalandeli. Data Center Knowledge. https://www.datacenterknowledge.com/hyperscalers/data-center-fire-google-suffers-electric-incident-3-abalandeli Claburn, T. (2023, April 26). I-Google Cloud ibandakanya eYurophu phakathi kwamanzi ebandayo, umoya. https://www.theregister.com/2023/04/26/google_cloud_outage/ Sverdlik, Y. (2021, March 9). Ukukhutshwa kwe-OVH ye-Strasbourg data center (SBG2). I-Data Center Knowledge. https://www.datacenterknowledge.com/uptime/fire-has-destroyed-ovh-s-strasbourg-data center-sbg2 Iindaba zeBloomberg. (2022, ngoJulayi 20). I-Google, i-Oracle data centers ifumaneka kwi-offline yeLondon. I-Data Center Knowledge. https://www.datacenterknowledge.com/cooling/google-oracle-data-centers-knocked-offline-by-london-heat Millward, W. (2024, 5 Disemba). I-10 iintlobo ezininzi ze-cloud ye-2024. CRN. https://www.crn.com/news/cloud/2024/the-10-biggest-cloud-outages-of-2024 Barrett, B. (2021, Juni 8). Indlela inkampani emangalisayo waya amaphuzu enkulu ye-internet. WIRED. https://www.wired.com/story/fastly-cdn-internet-outages-2021/

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

Le audio iveliswe ngolwimi lokuqala lwebali!

Inde kakhulu; Ukufunda

Make resilience your competitive advantage

I-Sia I-Redefines I-Cloud Reliability nge-Performance Continuous by Design

About Author

IZIMVO

ZIJONGE IIMPAWU

ELI NQAKU LINIKEZELWE KU

Related Stories

Light-Mode

Classic

Newspaper

Dark-Mode

Neon Noir

Minty

HN StartUps