Aloha mai e ka poe heluhelu Habr. Me kēia ʻatikala e wehe mākou i kahi moʻolelo e kamaʻilio e pili ana i ka ʻōnaehana hyperconverged AERODISK vAIR a mākou i kūkulu ai. I ka hoʻomaka, makemake mākou e haʻi i nā mea āpau e pili ana i nā mea āpau ma ka ʻatikala mua, akā paʻakikī loa ka ʻōnaehana, no laila e ʻai mākou i ka ʻelepani ma nā ʻāpana.
E hoʻomaka kākou i ka moʻolelo me ka mōʻaukala o ka hana ʻana o ka ʻōnaehana, e komo i loko o ka ʻōnaehana faila ARDFS, ʻo ia ke kumu o vAIR, a kamaʻilio liʻiliʻi hoʻi e pili ana i ke kūlana o kēia hopena ma ka mākeke Lūkini.
Ma nā ʻatikala e hiki mai ana, e kamaʻilio mākou i nā kikoʻī e pili ana i nā ʻāpana hoʻolālā like ʻole (cluster, hypervisor, load balancer, monitoring system, etc.), ke kaʻina hana hoʻonohonoho, hoʻāla i nā pilikia laikini, hōʻike ʻokoʻa i nā hoʻokolohua crash a, ʻoiaʻiʻo, e kākau e pili ana i ka hoʻāʻo ukana a nui ana. E hāʻawi pū mākou i kahi ʻatikala kaʻawale i ka mana kaiāulu o vAIR.
He moʻolelo anei ʻo Aerodisk e pili ana i nā ʻōnaehana mālama? A i ʻole no ke aha mākou i hoʻomaka ai e hana i ka hyperconvergence ma kahi mua?
I ka wā mua, hiki mai ka manaʻo e hana i kā mākou hyperconvergence ponoʻī iā mākou ma kahi o 2010. I kēlā manawa, ʻaʻohe Aerodisk a i ʻole nā ʻoluʻolu like (nā ʻōnaehana boxed hyperconverged ʻoihana) ma ka mākeke. ʻO kā mākou hana penei: mai kahi hoʻonohonoho o nā kikowaena me nā disks kūloko, i hui pū ʻia e kahi interconnect ma o ka protocol Ethernet, pono ia e hana i kahi mālama lōʻihi a hoʻomaka i nā mīkini virtual a me kahi pūnaewele polokalamu ma laila. Pono e hoʻokō ʻia kēia mau mea me ka ʻole o nā ʻōnaehana mālama (no ka mea ʻaʻohe kālā no nā ʻōnaehana mālama a me kāna lako, a ʻaʻole mākou i haku i kā mākou ʻōnaehana mālama ponoʻī).
Ua ho'āʻo mākou i nā hāʻina open source a hoʻopau hope i kēia pilikia, akā paʻakikī loa ka hopena a paʻakikī ke hana hou. Ma waho aʻe, aia kēia hoʻonā i ka ʻāpana o "Ke hana nei? Mai hoopa aku! No laila, i ka hoʻoponopono ʻana i kēlā pilikia, ʻaʻole mākou i hoʻomohala hou i ka manaʻo o ka hoʻololi ʻana i ka hopena o kā mākou hana i kahi huahana piha.
Ma hope o kēlā hanana, ua neʻe mākou mai kēia manaʻo, akā naʻe ua loaʻa iā mākou ka manaʻo ua hiki ke hoʻopau ʻia kēia pilikia, a ʻoi aku ka maikaʻi o ia ʻano hopena. Ma hope iho, ua hōʻoia wale nā huahana HCI o nā hui haole i kēia manaʻo.
No laila, i ka waena o 2016, ua hoʻi mākou i kēia hana ma ke ʻano he ʻāpana o ka hana ʻana i kahi huahana piha. I kēlā manawa ʻaʻohe o mākou pilina me nā mea hoʻopukapuka, no laila pono mākou e kūʻai i kahi kūʻai hoʻomohala no kā mākou kālā ʻaʻole nui loa. Ma hope o ka hōʻiliʻili ʻana i nā kikowaena hoʻohana a hoʻololi iā Avito, ua iho mākou i ka ʻoihana.
ʻO ka hana mua, ʻo ia ka hana ʻana i kā mākou ponoʻī, akā maʻalahi, akā i kā mākou ʻōnaehana faila ponoʻī, hiki ke hoʻokaʻawale i nā ʻikepili ma ke ʻano o nā poloka virtual ma ka helu nth o nā nodes cluster, i hoʻopili ʻia e kahi interconnect ma o Ethernet. I ka manawa like, pono ka FS i ka nui a me ka maʻalahi a kūʻokoʻa i nā ʻōnaehana pili, ʻo ia hoʻi. e hoʻokaʻawale ʻia mai vAIR ma ke ʻano o "kahi hale mālama wale nō".
Manao vAIR mua
Ua haʻalele mākou i ka hoʻohana ʻana i nā ʻōnaehana open source i mākaukau no ka hoʻonohonoho ʻana i ka waiho ʻana (ceph, gluster, luster a me nā mea like) no kā mākou hoʻomohala ʻana, no ka mea, ua loaʻa iā mākou ka nui o ka ʻike papahana me lākou. ʻOiaʻiʻo, ʻoi aku ka maikaʻi o kēia mau hoʻonā, a ma mua o ka hana ʻana ma Aerodisk, ua hoʻokō mākou i ʻoi aku ma mua o hoʻokahi papahana hoʻohui me lākou. Akā, hoʻokahi mea e hoʻokō i kahi hana kūikawā no ka mea kūʻai aku, hoʻomaʻamaʻa limahana a, malia paha, kūʻai i ke kākoʻo o kahi mea kūʻai aku nui, a he mea ʻē aʻe e hana i kahi huahana maʻalahi e hoʻohana ʻia no nā hana like ʻole, a mākou, ma ke ʻano he mea kūʻai aku, ʻike paha e pili ana iā mākou iho ʻaʻole mākou e hana. No ke kumu ʻelua, ʻaʻole kūpono nā huahana open source no mākou, no laila ua hoʻoholo mākou e hana i kahi ʻōnaehana faila i hāʻawi ʻia iā mākou iho.
ʻElua mau makahiki ma hope mai, ua loaʻa i kekahi mau mea hoʻomohala (nāna i hoʻohui i ka hana ma vAIR me ka hana ma ka ʻōnaehana mālama ʻenehana maʻamau) i kahi hopena.
Ma ka 2018, ua kākau mākou i kahi ʻōnaehana faila maʻalahi a hoʻohui ʻia me ka lako pono. Ua hoʻohui ka ʻōnaehana i nā disks kino (local) mai nā kikowaena like ʻole i loko o hoʻokahi ʻauʻau palahalaha ma o kahi pilina kūloko a "ʻoki" iā lākou i loko o nā poloka virtual, a laila ua hoʻokumu ʻia nā mea hana me nā pae like ʻole o ka hoʻomanawanui hewa mai nā poloka virtual, kahi i hana ʻia ai nā mea virtual. a hoʻokō ʻia me nā kaʻa KVM hypervisor.
ʻAʻole mākou i pilikia nui i ka inoa o ka ʻōnaehana faila a kapa ʻia ʻo ia ʻo ARDFS (manaʻo i ke ʻano o ia mea))
Ua nānā maikaʻi kēia prototype (ʻaʻole i ʻike ʻia, ʻoiaʻiʻo, ʻaʻohe hoʻolālā hiʻohiʻona) a hōʻike i nā hopena maikaʻi ma ke ʻano o ka hana a me ka scaling. Ma hope o ka hopena maoli mua, ua hoʻonohonoho mākou i kēia papahana i ka neʻe ʻana, e hoʻonohonoho ana i kahi kaiapuni hoʻomohala piha a me kahi hui ʻokoʻa i hana wale me vAIR.
I kēlā manawa wale nō, ua oʻo ka hoʻolālā nui o ka hopena, ʻaʻole i hoʻololi nui ʻia.
Luʻu i loko o ka ʻōnaehana faila ARDFS
ʻO ARDFS ke kumu o ka vAIR, ka mea e hāʻawi ana i ka mālama ʻana i ka ʻikepili i hāʻawi ʻia, ʻae i ka hewa ma waena o ka pūʻulu holoʻokoʻa. ʻO kekahi o (akā ʻaʻole wale) nā hiʻohiʻona o ARDFS ʻaʻole ia e hoʻohana i nā kikowaena hoʻolaʻa hou no ka metadata a me ka hoʻokele. Ua hoʻokumu mua ʻia kēia e hoʻomaʻamaʻa i ka hoʻonohonoho ʻana o ka hoʻonā a no kona hilinaʻi.
Hoʻolālā waihona
I loko o nā node a pau o ka pūʻulu, hoʻonohonoho ʻo ARDFS i kahi wai kūpono mai nā wahi disk āpau i loaʻa. He mea nui ka hoʻomaopopo ʻana ʻaʻole he ʻikepili a i ʻole i hoʻopaʻa ʻia kahi loko, akā markup wale nō, ʻo ia. ʻO nā node me ka vAIR i hoʻokomo ʻia, ke hoʻohui ʻia i ka pūʻulu, e hoʻohui ʻia i ka pūnāwai ARDFS kaʻana like a me nā kumuwaiwai disk e kaʻana like ʻia ma ka pūʻulu holoʻokoʻa (a loaʻa no ka mālama ʻikepili e hiki mai ana). Hāʻawi kēia ala iā ʻoe e hoʻohui a wehe i nā nodes ma ka lele me ka ʻole o ka hopena koʻikoʻi i ka ʻōnaehana holo mua. ʻO kēlā mau mea. He mea maʻalahi loa ka ʻōnaehana "ma nā pōhaku lepo", hoʻohui a wehe paha i nā nodes i ka hui inā pono.
Hoʻohui ʻia nā disks virtual (nā mea mālama no nā mīkini virtual) ma luna o ka loko ARDFS, i kūkulu ʻia mai nā poloka virtual o 4 megabytes ka nui. Hoʻopaʻa pololei nā disks virtual i ka ʻikepili. Hoʻonohonoho pū ʻia ka ʻōnaehana hoʻomanawanui hewa ma ka pae diski virtual.
E like me kāu i manaʻo mua ai, no ka hoʻomanawanui hewa o ka subsystem disk, ʻaʻole mākou e hoʻohana i ka manaʻo o RAID (Redundant array of independent Disks), akā hoʻohana i ka RAIN (Redundant array of independent Nodes). ʻO kēlā mau mea. Ana ʻia ka ʻae ʻana i ka hewa, automated, a mālama ʻia ma muli o nā nodes, ʻaʻole nā disks. ʻO nā disks, ʻoiaʻiʻo, he mea mālama ia, ʻo lākou, e like me nā mea ʻē aʻe, ke nānā ʻia, hiki iā ʻoe ke hana i nā hana maʻamau āpau me lākou, me ka hui pū ʻana i kahi RAID hāmeʻa kūloko, akā e hana pono ka cluster ma nā nodes.
Ma kahi kūlana āu e makemake nui ai iā RAID (no ka laʻana, kahi hiʻohiʻona e kākoʻo ana i nā hemahema he nui ma nā puʻupuʻu liʻiliʻi), ʻaʻohe mea e pale iā ʻoe mai ka hoʻohana ʻana i nā mea hoʻokele RAID kūloko, a me ke kūkulu ʻana i kahi waiho waiho ʻia a me kahi hoʻolālā RAIN ma luna. Ke ola nei kēia ʻano hiʻohiʻona a kākoʻo ʻia e mākou, no laila e kamaʻilio mākou e pili ana iā ia ma kahi ʻatikala e pili ana i nā hiʻohiʻona maʻamau no ka hoʻohana ʻana i vAIR.
ʻO nā papahana hoʻomanawanui hewa ʻana
Hiki ke loaʻa i ʻelua papa hana hoʻomanawanui no nā disks virtual ma vAIR:
1) Ka helu hoʻopiʻi a i ʻole ka hana hou ʻana - maʻalahi kēia ʻano o ka hoʻomanawanui hewa e like me ka lāʻau a me ke kaula. Hana ʻia ka hoʻopili like ʻana ma waena o nā node me ka helu o 2 (2 kope no kēlā me kēia hui) a i ʻole 3 (3 kope, kēlā me kēia). ʻAe ʻo RF-2 i kahi disk virtual e kū i ka hāʻule ʻole o hoʻokahi node i ka hui, akā "ʻai" i ka hapalua o ka leo pono, a ʻo RF-3 e kū i ka hiki ʻole o 2 nodes i ka hui, akā mālama ʻia 2/3 o ka puʻupuʻu. leo pono no kona pono. Ua like loa kēia kumumanaʻo me RAID-1, ʻo ia hoʻi, kahi disk virtual i hoʻonohonoho ʻia ma RF-2 e kūʻē i ka hemahema o kekahi node i loko o ka pūpū. I kēia hihia, e maikaʻi nā mea a pau me ka ʻikepili a ʻaʻole e pau ka I/O. Ke hoʻi ka node hāʻule i ka lawelawe, e hoʻomaka ka hoʻihoʻi / hoʻonohonoho ʻana i ka ʻikepili.
Aia ma lalo nā hiʻohiʻona o ka puʻunaue ʻana o ka ʻikepili RF-2 a me RF-3 ma ke ʻano maʻamau a i kahi kūlana hemahema.
Loaʻa iā mākou kahi mīkini virtual me ka mana o 8MB o ka ʻikepili kūikawā (pono), e holo ana ma 4 vAIR nodes. ʻIke ʻia i ka ʻoiaʻiʻo ʻaʻole hiki ke loaʻa kahi leo liʻiliʻi, akā no kahi hoʻolālā e hōʻike ana i ka loiloi o ka hana ARDFS, ʻo kēia hiʻohiʻona ka mea maopopo loa. ʻO AB he 4MB mau poloka virtual i loaʻa nā ʻikepili mīkini virtual kū hoʻokahi. Hana ʻo RF-2 i ʻelua kope o kēia mau poloka A1+A2 a me B1+B2. "Hoʻomoe ʻia" kēia mau poloka ma nā node, e pale ana i ka hui ʻana o ka ʻikepili like ma ka node hoʻokahi, ʻo ia hoʻi, ʻaʻole e loaʻa ke kope A1 ma ka node like me ke kope A2. Hoʻokahi me B1 a me B2.
Inā hāʻule kekahi o nā node (no ka laʻana, node No. 3, aia kahi kope o B1), hoʻāla ʻia kēia kope ma ka node kahi i loaʻa ʻole kope o kāna kope (ʻo ia hoʻi, he kope o B2).
No laila, hiki i ka diski virtual (a me ka VM, e like me ia) ke ola maʻalahi i ka hāʻule ʻole o hoʻokahi node i ka papahana RF-2.
ʻO ka papahana replication, ʻoiai maʻalahi a hilinaʻi hoʻi, loaʻa i ka pilikia like me RAID1 - ʻaʻole lawa ka wahi hoʻohana.
2) Erasure coding or erasure coding (ʻike pū ʻia ʻo “redundant coding”, “erasure coding” a i ʻole “redundancy code”) e hoʻoponopono i ka pilikia ma luna. ʻO EC kahi hoʻolālā redundancy e hāʻawi ana i ka loaʻa o ka ʻikepili kiʻekiʻe me ka haʻahaʻa o ka disk space overhead e hoʻohālikelike ʻia me ka hana hou ʻana. ʻO ka loina hana o kēia mīkini e like me RAID 5, 6, 6P.
I ka hoʻopaʻa ʻana, hoʻokaʻawale ke kaʻina hana EC i kahi poloka virtual (4MB ma ke ʻano maʻamau) i loko o kekahi mau "ʻokoʻa ʻikepili" liʻiliʻi ma muli o ke kumumanaʻo EC (no ka laʻana, hoʻokaʻawale ka papahana 2+1 i kēlā me kēia poloka 4MB i nā ʻāpana 2 2MB). A laila, hoʻopuka kēia kaʻina hana i nā "parity chunks" no nā "chunk data" ʻaʻole i ʻoi aku ka nui ma mua o kekahi o nā ʻāpana i māhele mua ʻia. I ka hoʻololi ʻana, hoʻopuka ʻo EC i nā ʻāpana i nalowale ma ka heluhelu ʻana i ka ʻikepili "ola" ma waena o ka pūʻulu holoʻokoʻa.
No ka laʻana, ʻo kahi diski virtual me kahi hoʻolālā 2 + 1 EC, i hoʻokō ʻia ma 4 mau puʻupuʻu puʻupuʻu, e pale maʻalahi i ka hāʻule ʻole o hoʻokahi node i ka pūʻulu e like me RF-2. I kēia hihia, e emi nā kumukūʻai o luna, ʻo ia hoʻi, ʻo ka coefficient hiki ke kūpono no RF-2 he 2, a no EC 2+1 ʻo ia ka 1,5.
No ka wehewehe ʻana i ka maʻalahi, ʻo ka mea nui ka māhele ʻana o ka poloka virtual i 2-8 (no ke aha mai 2 a 8, e ʻike i lalo) "nā ʻāpana", a no kēia mau ʻāpana "'āpana" o ka parity o kahi leo like.
ʻO ka hopena, ua puʻunaue like ʻia ka ʻikepili a me ka parity ma nā node āpau o ka pūʻulu. I ka manawa like, e like me ka replication, ARDFS e puʻunaue aunoa i ka ʻikepili ma waena o nā node i ke ʻano e pale ai i ka ʻikepili like (kope o ka ʻikepili a me kā lākou parity) mai ka mālama ʻana ma ka node hoʻokahi, i mea e hoʻopau ai i ka manawa o ka nalowale ʻana o ka ʻikepili ma muli o i ka ʻoiaʻiʻo e hoʻopau koke ʻia ka ʻikepili a me kā lākou parity ma kahi node mālama i hāʻule ʻole.
Aia ma lalo kahi laʻana, me ka mīkini virtual 8 MB like a me 4 nodes, akā me kahi hoʻolālā EC 2+1.
Ua māhele ʻia nā poloka A a me B i ʻelua ʻāpana o 2 MB kēlā me kēia (ʻelua no ka 2+1), ʻo ia hoʻi, A1+A2 a me B1+B2. ʻAʻole like me ka replica, ʻaʻole ʻo A1 he kope o A2, ʻo ia ka poloka virtual A, i māhele ʻia i ʻelua mau ʻāpana, like me ka poloka B. Ma ka huina, loaʻa iā mākou ʻelua pūʻulu o 4MB, ʻo kēlā me kēia me ʻelua mau ʻāpana ʻelua MB. A laila, no kēlā me kēia o kēia mau pūʻulu, ua helu ʻia ka parity me ka nui o ka ʻāpana ʻaʻole i ʻoi aku ma mua o hoʻokahi ʻāpana (ʻo ia hoʻi 2 MB), loaʻa iā mākou he + 2 mau ʻāpana parity (AP a me BP). Loaʻa iā mākou ka 4 × 2 data + 2 × 2 parity.
A laila, "kau ʻia" nā ʻāpana i waena o nā nodes i ʻole e hui pū ka ʻikepili me ko lākou parity. ʻO kēlā mau mea. ʻAʻole ʻo A1 a me A2 ma ka node like me AP.
I ka wā o ka hāʻule ʻole o hoʻokahi node (no ka laʻana, ʻo ke kolu hoʻi), e hoʻihoʻi ʻia ka pahu hāʻule B1 mai ka parity BP, i mālama ʻia ma ka node No. ʻaʻohe B-parity, ʻo ia hoʻi. apana BP. Ma kēia laʻana, ʻo kēia ka node No. 2
Manaʻo wau he nīnau ka mea heluhelu:
"ʻO nā mea a pau āu i wehewehe ai ua hoʻokō lōʻihi ʻia e nā mea hoʻokūkū a me nā hoʻonā open source, he aha ka ʻokoʻa ma waena o kāu hoʻokō ʻana o EC ma ARDFS?"
A laila e loaʻa nā hiʻohiʻona hoihoi o ARDFS.
Holoi i ka coding me ka nānā ʻana i ka maʻalahi
I ka hoʻomaka ʻana, ua hāʻawi mākou i kahi hoʻolālā EC X+Y maʻalahi, kahi e like ai ʻo X me kahi helu mai ka 2 a hiki i ka 8, a ua like ʻo Y me kahi helu mai ka 1 a hiki i ka 8, akā ʻoi aku ka liʻiliʻi a i ʻole like me X. no ka maʻalahi. ʻO ka hoʻonui ʻana i ka helu o nā ʻāpana ʻikepili (X) kahi i hoʻokaʻawale ʻia ai ka poloka virtual e hiki ai ke hōʻemi i nā kumukūʻai ma luna, ʻo ia hoʻi, ka hoʻonui ʻana i kahi kūpono.
ʻO ka hoʻonui ʻana i ka helu o nā parity chunks (Y) e hoʻonui i ka hilinaʻi o ka disk virtual. ʻOi aku ka nui o ka waiwai Y, ʻoi aku ka nui o nā nodes i ka pūʻulu hiki ke hāʻule. ʻOiaʻiʻo, ʻo ka hoʻonui ʻana i ka nui parity e hōʻemi i ka nui o ka hiki ke hoʻohana, akā he kumukūʻai kēia e uku no ka hilinaʻi.
ʻO ka hilinaʻi o ka hana ma nā kaapuni EC ua aneane pololei: ʻoi aku ka nui o nā "'āpana", ʻoi aku ka haʻahaʻa o ka hana; ma ʻaneʻi, ʻoiaʻiʻo, pono kahi ʻike kaulike.
ʻO kēia ala e hiki ai i nā luna hoʻoponopono ke hoʻonohonoho i ka waiho ʻana me ka maʻalahi loa. I loko o ka loko ARDFS, hiki iā ʻoe ke hoʻohana i nā hoʻolālā hoʻomanawanui hewa a me kā lākou hui pū ʻana, i ko mākou manaʻo, he mea pono loa.
Aia ma lalo kahi papa e hoʻohālikelike ana i kekahi mau (ʻaʻole hiki ke hiki) RF a me EC.
Hōʻike ka papa ʻaina i ka hui pū ʻana o "terry" EC 8 + 7, e ʻae ai i ka nalowale ʻana o 7 nodes i loko o kahi puʻupuʻu i ka manawa like, "ʻai" liʻiliʻi ka hoʻohana ʻana (1,875 versus 2) ma mua o ka replication maʻamau, a pale i nā manawa 7 maikaʻi. , ʻo ia ka mea e hana ai i kēia ʻōnaehana pale, ʻoiai ʻoi aku ka paʻakikī, ʻoi aku ka maikaʻi i nā kūlana kahi e pono ai e hōʻoia i ka hilinaʻi kiʻekiʻe i nā kūlana o ka palena diski. I ka manawa like, pono ʻoe e hoʻomaopopo i kēlā me kēia "hoʻohui" iā X a i ʻole Y e lilo i mea hana hou ma luna, no laila i ka triangle ma waena o ka hilinaʻi, mālama a me ka hana pono ʻoe e koho pono. No kēia kumu, e hāʻawi mākou i kahi ʻatikala ʻokoʻa e holoi i ka nui coding.
ʻO ka hilinaʻi a me ke kūʻokoʻa o ka ʻōnaehana faila
Holo ʻo ARDFS ma ka ʻāina ma nā node āpau o ka pūʻulu a hoʻonohonoho iā lākou me ka hoʻohana ʻana i kāna mau ala ponoʻī ma o nā kikowaena Ethernet i hoʻolaʻa ʻia. ʻO ka mea koʻikoʻi, ʻo ka ARDFS kūʻokoʻa i ka synchronize ʻaʻole wale i ka ʻikepili, akā ʻo ka metadata pū kekahi e pili ana i ka waiho ʻana. ʻOiai mākou e hana ana ma ARDFS, ua aʻo mākou i kekahi o nā hopena i loaʻa a ʻike mākou he nui ka hoʻonohonoho ʻana i ka metadata file me ka hoʻohana ʻana i kahi DBMS puʻupuʻu i waho, a mākou e hoʻohana ai no ka synchronization, akā hoʻonohonoho wale nō, ʻaʻole FS metadata (e pili ana i kēia a me nā subsystem e pili ana. ma ka ʻatikala aʻe).
ʻO ka hoʻonohonoho ʻana i ka metadata FS me ka hoʻohana ʻana i kahi DBMS o waho, ʻoiaʻiʻo, he hopena hana, akā ʻo ke kūlike o ka ʻikepili i mālama ʻia ma ARDFS e hilinaʻi ʻia i ka DBMS waho a me kāna ʻano (a, ʻo ka ʻōlelo pololei, he wahine capricious), i loko ino ko makou manao. No ke aha mai? Inā pōʻino ka metadata FS, hiki ke ʻōlelo ʻia ka ʻikepili FS ponoʻī "aloha," no laila ua hoʻoholo mākou e lawe i kahi ala paʻakikī akā hilinaʻi.
Ua hana mākou i ka metadata synchronization subsystem no ARDFS iā mākou iho, a ke noho kūʻokoʻa nei ia mai nā subsystem e pili ana. ʻO kēlā mau mea. ʻaʻole hiki i nā subsystem ʻē aʻe ke hōʻino i ka ʻikepili ARDFS. I ko mākou manaʻo, ʻo kēia ke ala hilinaʻi a pololei, akā na ka manawa e haʻi inā he ʻoiaʻiʻo kēia. Eia kekahi, aia kekahi pono hou me kēia ala. Hiki ke hoʻohana kūʻokoʻa ʻo ARDFS me ka vAIR, e like me ka waiho ʻana, e hoʻohana maoli mākou i nā huahana e hiki mai ana.
ʻO ka hopena, ma ka hoʻomohala ʻana i ka ARDFS, ua loaʻa iā mākou kahi ʻōnaehana faila maʻalahi a hilinaʻi e hāʻawi i kahi koho e hiki ai iā ʻoe ke mālama i ka hiki a hāʻawi i nā mea āpau i ka hana, a i ʻole e hana i kahi waihona hilinaʻi nui loa ma ke kumukūʻai kūpono, akā e hōʻemi ana i nā koi hana.
Me kahi kulekele laikini maʻalahi a me kahi hiʻohiʻona hāʻawi maʻalahi (e nānā ana i mua, ua laikini ʻia ka vAIR e ka node, a hāʻawi ʻia ma ke ʻano he polokalamu a i ʻole he pūʻolo polokalamu), hiki iā ʻoe ke hoʻoponopono pololei i ka hopena i nā ʻano mea kūʻai aku. a laila mālama maʻalahi i kēia kaulike.
ʻO wai ka mea e pono ai kēia hana mana?
Ma ka ʻaoʻao hoʻokahi, hiki iā mākou ke ʻōlelo aku aia nā mea pāʻani ma ka mākeke i loaʻa nā hopena koʻikoʻi i ke kahua o ka hyperconvergence, a ʻo kēia kahi a mākou e hele nei. Me he mea la he oiaio keia olelo, AKA...
Ma ka ʻaoʻao ʻē aʻe, i ka wā e hele ai mākou i nā kula a kamaʻilio me nā mea kūʻai aku, ʻike mākou a me kā mākou mau hoa ʻaʻole kēia ka hihia. Nui nā hana no ka hyperconvergence, ma kekahi mau wahi ʻaʻole ʻike ka poʻe i ke kū ʻana o ia mau hopena, i nā mea ʻē aʻe he kumukūʻai, i nā poʻe ʻē aʻe he mau hoʻokolohua kūleʻa ʻole o nā hopena ʻē aʻe, a ma nā wahi ʻē aʻe ua pāpā lākou i ke kūʻai ʻana ma muli o nā hoʻopaʻi. Ma keʻano laulā, ua lilo ka mahinaʻai i ka palau ʻole, no laila ua hele mākou e hānai i ka lepo puʻupaʻa))).
ʻOi aku ka maikaʻi o ka ʻōnaehana mālama ma mua o GCS?
Ke hana nei mākou me ka mākeke, nīnau pinepine ʻia mākou i ka wā e ʻoi aku ka maikaʻi o ka hoʻohana ʻana i kahi papahana maʻamau me nā ʻōnaehana mālama, a i ka wā e hoʻohana ai i ka hyperconvergent? He nui nā hui e hana ana i ka GCS (ʻoi aku ka poʻe i loaʻa ʻole nā pūnaewele mālama i kā lākou kōpili) e ʻōlelo: "Ke lilo nei nā ʻōnaehana mālama i ka wā kahiko, hyperconverged wale nō!" He ʻōlelo wiwo ʻole kēia, akā ʻaʻole ia e hōʻike piha i ka ʻoiaʻiʻo.
I ka ʻoiaʻiʻo, ke neʻe nei ka mākeke mālama i ka hyperconvergence a me nā hopena like, akā aia mau ka "akā".
ʻO ka mea mua, ʻaʻole hiki ke kūkulu hou ʻia nā kikowaena data a me nā ʻōnaehana IT i kūkulu ʻia e like me ka hoʻolālā maʻamau me nā ʻōnaehana mālama, no laila ʻo ka hoʻololi ʻana a me ka hoʻopau ʻana o ia mau ʻōnaehana he hoʻoilina no 5-7 mau makahiki.
ʻO ka lua, ʻo ka ʻōnaehana i kūkulu ʻia i kēia manawa no ka hapa nui (ʻo ia hoʻi ka Russian Federation) i kūkulu ʻia e like me ka papa hana maʻamau me ka hoʻohana ʻana i nā ʻōnaehana mālama, ʻaʻole no ka ʻike ʻole o ka poʻe e pili ana i ka hyperconvergence, akā no ka mea he hou ka mākeke hyperconvergence. ʻaʻole i hoʻokumu ʻia nā kūlana , ʻaʻole i hoʻomaʻamaʻa ʻia ka poʻe IT, he liʻiliʻi ko lākou ʻike, akā pono lākou e kūkulu i nā kikowaena data ma aneʻi a i kēia manawa. A e mau ana kēia ʻano no kekahi 3-5 mau makahiki (a laila hoʻoilina hou, e ʻike i ka helu 1).
ʻO ke kolu, aia kahi palena ʻenehana maʻemaʻe i nā liʻiliʻi liʻiliʻi o 2 milliseconds no ke kākau ʻana (me ka ʻole o ka cache kūloko, ʻoiaʻiʻo), ʻo ia ke kumukūʻai o ka mālama ʻana.
ʻAe, mai poina kākou e pili ana i ka hoʻohana ʻana i nā kikowaena kino nui e makemake ana i ka scaling vertical o ka subsystem disk.
Nui nā hana pono a kaulana i ʻoi aku ka maikaʻi o nā ʻōnaehana mālama ma mua o GCS. Maʻaneʻi,ʻoiaʻiʻo,ʻaʻole eʻae like nā mea hana i loaʻaʻole nā pūnaewele mālama i kā lākou waihona huahana, akā ua mākaukau mākou e hoʻopaʻapaʻa kūpono. ʻOiaʻiʻo, ʻo mākou, ma ke ʻano he mea hoʻomohala o nā huahana ʻelua, e hoʻohālikelike maoli i nā ʻōnaehana mālama a me GCS i kekahi o kā mākou puke e hiki mai ana, kahi e hōʻike pono ai mākou i ka mea ʻoi aku ka maikaʻi ma lalo o nā kūlana.
A ma hea kahi e hana maikaʻi ai nā hoʻonā hyperconverged ma mua o nā ʻōnaehana mālama?
Ma muli o nā manaʻo ma luna, hiki ke huki ʻia ʻekolu mau hopena maopopo:
- Ma kahi o 2 milliseconds o ka latency no ka hoʻopaʻa ʻana, e kū mau ana i kekahi huahana (i kēia manawa ʻaʻole mākou e kamaʻilio e pili ana i nā synthetics, hiki ke hōʻike ʻia nā nanoseconds ma nā synthetics), ʻaʻole pono, kūpono ka hyperconvergent.
- Ma kahi e hiki ai ke hoʻololi i ka ukana mai nā kikowaena kino nui i mau mea liʻiliʻi liʻiliʻi a puʻunaue ʻia i waena o nā nodes, e hana maikaʻi nō hoʻi ka hyperconvergence ma laila.
- Inā ʻoi aku ka nui o ka scaling horizontal ma mua o ka scaling vertical, e hana maikaʻi ʻo GCS ma laila.
He aha kēia mau hoʻonā?
- ʻO nā lawelawe ʻoihana maʻamau āpau (lawelawe kuhikuhi, leka uila, EDMS, nā kikowaena faila, nā ʻōnaehana ERP liʻiliʻi a i ʻole nā BI, etc.). Kapa mākou i kēia "helu helu nui".
- ʻO ka ʻōnaehana o nā mea hoʻolako kapua, kahi e pono ai e wikiwiki a hoʻohālikelike i ka hoʻonui ākea ākea a maʻalahi "ʻoki" i ka nui o nā mīkini virtual no nā mea kūʻai aku.
- Virtual desktop infrastructure (VDI), kahi e holo ai ka nui o na mīkini virtual mea hoʻohana liʻiliʻi a "lele" mālie i loko o kahi hui like.
- ʻO nā pūnaewele lālā, kahi e pono ai kēlā me kēia lālā i kahi kūlana maʻamau, hewa-ʻae, akā ʻaʻohe kumu kūʻai o nā mīkini virtual 15-20.
- ʻO kēlā me kēia helu helu i hāʻawi ʻia (nā lawelawe ʻikepili nui, no ka laʻana). Kahi e hele ai ka ukana ʻaʻole "hohonu", akā "ka laulā".
- ʻO nā wahi hoʻāʻo kahi e ʻae ʻia ai nā lohi liʻiliʻi, akā aia nā palena kālā, no ka mea he mau hoʻāʻo kēia.
I kēia manawa, no kēia mau hana mākou i hana ai iā AERODISK vAIR a ma luna o lākou kā mākou e kālele nei (holomua i kēia manawa). E loli koke ana paha kēia, no ka mea... ʻaʻole kū mālie ka honua.
No laila…
Hoʻopau kēia i ka ʻāpana mua o kahi ʻatikala nui; ma ka ʻatikala aʻe e kamaʻilio mākou e pili ana i ka hoʻolālā o ka hopena a me nā mea i hoʻohana ʻia.
Aloha mākou i nā nīnau, nā manaʻo a me nā hoʻopaʻapaʻa maikaʻi.
Source: www.habr.com