{"id":40729,"date":"2020-02-03T14:42:16","date_gmt":"2020-02-03T11:42:16","guid":{"rendered":"https:\/\/prohoster.info\/blog\/blog_prohoster\/kratkoe-sravnenie-arhitektury-sds-ili-poisk-podhodyashhej-platformy-hraneniya-glustervscephvsvirtuozzostorage"},"modified":"2020-02-03T14:42:16","modified_gmt":"2020-02-03T11:42:16","slug":"kratkoe-sravnenie-arhitektury-sds-ili-poisk-podhodyashhej-platformy-hraneniya-glustervscephvsvirtuozzostorage","status":"publish","type":"post","link":"https:\/\/prohoster.info\/ro\/blog\/kratkoe-sravnenie-arhitektury-sds-ili-poisk-podhodyashhej-platformy-hraneniya-glustervscephvsvirtuozzostorage","title":{"rendered":"O compara\u021bie scurt\u0103 a arhitecturii SDS sau c\u0103utarea unei platforme de stocare potrivite (GlusterVsCephVsVirtuozzoStorage)","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"<p>Acest articol este scris pentru a ajuta cititorul s\u0103 aleag\u0103 solu\u021bia potrivit\u0103 \u0219i s\u0103 \u00een\u021beleag\u0103 diferen\u021bele dintre diversele SDS, cum ar fi Gluster, Ceph \u0219i Vstorage (Virtuozzo).<\/p>\n<p>\u00cen text se folosesc linkuri c\u0103tre articole cu o expunere mai detaliat\u0103 a anumitor probleme, a\u0219a c\u0103 descrierile vor fi c\u00e2t mai concise, folosind doar punctele cheie, f\u0103r\u0103 informa\u021bii inutile \u0219i introductive pe care le pute\u021bi descoperi pe cont propriu \u00een spa\u021biul online. <\/p>\n<p>De fapt, subiectele abordate necesit\u0103 un ton specific al textului, dar \u00een lumea modern\u0103 tot mai mul\u021bi oameni nu doresc s\u0103 citeasc\u0103 mult))), a\u0219a c\u0103 poate fi citit pe scurt pentru a face o alegere, iar dac\u0103 e ceva neclar, pot fi explorate linkurile sau c\u0103utate cuvintele necunoscute))), iar acest articol este ca o acoperire transparent\u0103 pentru aceste subiecte profunde, ar\u0103t\u00e2nd con\u021binutul \u2013 principalele puncte cheie ale fiec\u0103rei solu\u021bii.<\/p>\n<h3>Gluster<\/h3>\n<p>\n\u00cencepem cu Gluster, care este folosit intens de c\u0103tre produc\u0103torii de platforme hiperconvergente cu SDS bazate pe open source pentru medii virtuale, \u0219i poate fi g\u0103sit pe site-ul RedHat la sec\u021biunea storage, unde se ofer\u0103 dou\u0103 op\u021biuni SDS: Gluster sau Ceph.<\/p>\n<p>Gluster const\u0103 dintr-un cadru de traduc\u0103tori \u2013 servicii care gestioneaz\u0103 toate sarcinile de distribuire a fi\u0219ierelor etc. Brick \u2013 serviciul care deserveste un disc, Volume \u2013 volum (pool) \u2013 care \u00eempreun\u0103 aceste brick-uri. Apoi urmeaz\u0103 un serviciu de distribuire a fi\u0219ierelor \u00een grupuri prin func\u021bia DHT (tabel de hash distribuit). Nu vom include serviciul de Sharding \u00een descriere, deoarece \u00een linkurile de mai jos va fi prezentat\u0103 descrierea problemelor legate de acesta.<\/p>\n<p><img decoding=\"async\" alt=\"O compara\u021bie scurt\u0103 a arhitecturii SDS sau c\u0103utarea unei platforme de stocare potrivite (GlusterVsCephVsVirtuozzoStorage)\" src=\"\/wp-content\/uploads\/2020\/02\/0c54588b5b09c9069ffbd0c5161c5e34.png\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<br \/>\nLa scriere, fi\u0219ierul este stocat integral \u00een brick \u0219i o copie a acestuia este scris\u0103 \u00een paralel pe brick-ul de pe serverul secundar. Ulterior, un al doilea fi\u0219ier va fi scris \u00een a doua grupare din dou\u0103 brick-uri (sau mai multe) de pe servere diferite.<br \/>\n<noindex><a rel=\"nofollow\" name=\"habracut\"><\/a><\/noindex> <br \/>\nDac\u0103 fi\u0219ierele au dimensiuni aproximativ egale \u0219i volumul va consta doar dintr-un singur grup, totul este \u00een regul\u0103, \u00eens\u0103 \u00een alte condi\u021bii, descrierile vor eviden\u021bia urm\u0103toarele probleme: <\/p>\n<ul>\n<li>spa\u021biul \u00een grupuri este folosit inegal, ceea ce depinde de dimensiunile fi\u0219ierelor, iar dac\u0103 \u00eentr-un grup nu exist\u0103 suficient loc pentru a scrie fi\u0219ierul \u2013 ve\u021bi ob\u021bine o eroare, fi\u0219ierul nu va fi scris \u0219i nu va fi redistribuit \u00een alt grup;<\/li>\n<li>\u00een timpul scrierii unui fi\u0219ier, IO se desf\u0103\u0219oar\u0103 doar pe un singur grup, celelalte st\u00e2nd inactive;<\/li>\n<li>nu se poate ob\u021bine IO pentru \u00eentregul volum \u00een timpul scrierii unui fi\u0219ier;<\/li>\n<li>\u0219i concep\u021bia general\u0103 pare mai pu\u021bin performant\u0103 din cauza lipsei distribuirii datelor pe blocuri, unde este mai simplu s\u0103 se realizeze un echilibru \u0219i s\u0103 se rezolve problema distribuirii uniforme, \u00een loc de cum este acum, fi\u0219ierul se plaseaz\u0103 \u00een brich integral.<\/li>\n<\/ul>\n<p>\nDin descrierea oficial\u0103 <noindex><a rel=\"nofollow\" href=\"https:\/\/docs.gluster.org\/en\/latest\/Quick-Start-Guide\/Architecture\/\">arhitecturii<\/a><\/noindex> se creeaz\u0103 involuntar \u00een\u021belegerea c\u0103 Gluster func\u021bioneaz\u0103 ca un stocare de fi\u0219iere deasupra unui RAID hardware clasic. Au existat \u00eencerc\u0103ri de a dezvolta fracturarea (Sharding) fi\u0219ierelor \u00een blocuri, dar totul este un supliment care impune pierderi de performan\u021b\u0103 la deja existentul abordare arhitectural\u0103, plus utilizarea unor componente liber distribuite cu limit\u0103ri \u00een performan\u021b\u0103 precum Fuse. Nu exist\u0103 servicii de metadate, ceea ce limiteaz\u0103 capacit\u0103\u021bile de performan\u021b\u0103 \u0219i rezilien\u021b\u0103 a stoc\u0103rii la distribuirea fi\u0219ierelor pe blocuri. Poate fi observat\u0103 o performan\u021b\u0103 mai bun\u0103 \u00een configura\u021bia \u201eDistributed Replicated\u201d \u0219i num\u0103rul de noduri trebuie s\u0103 fie de cel pu\u021bin 6 pentru a organiza o replicare fiabil\u0103 3 cu o distribu\u021bie optim\u0103 a sarcinii.<\/p>\n<p>Aceste concluzii sunt de asemenea legate de descrierea experien\u021bei utiliz\u0103rii <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/company\/croccloudteam\/blog\/353666\/\">Gluster<\/a><\/noindex> \u0219i la compararea cu <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/company\/croccloudteam\/blog\/430474\/\">Ceph<\/a><\/noindex>, precum \u0219i este o descriere a experien\u021bei \u00een a ajunge la o \u00een\u021belegere a acestei configura\u021bii mai performante \u0219i mai fiabile <noindex><a rel=\"nofollow\" href=\"https:\/\/www.linux.org.ru\/forum\/general\/13316001\">\u201eReplicated Distributed\u201d.<\/a><\/noindex><br \/>\n<img decoding=\"async\" alt=\"O compara\u021bie scurt\u0103 a arhitecturii SDS sau c\u0103utarea unei platforme de stocare potrivite (GlusterVsCephVsVirtuozzoStorage)\" src=\"\/wp-content\/uploads\/2020\/02\/6868ddf69e18bec0b50ae5aeba74440d.png\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<br \/>\n\u00cen imagine este prezentat\u0103 distribu\u021bia sarcinii la scrierea a dou\u0103 fi\u0219iere, unde copiile primului fi\u0219ier sunt distribuite pe primele trei servere, care sunt grupate \u00een grupul volume 0, iar trei copii ale celui de-al doilea fi\u0219ier se plaseaz\u0103 \u00een al doilea grup volume 1 din trei servere. Fiecare server are un disc.<\/p>\n<p>Concluzia general\u0103 este c\u0103 Gluster poate fi utilizat, dar cu \u00een\u021belegerea c\u0103 vor exista limit\u0103ri \u00een performan\u021b\u0103 \u0219i rezilien\u021b\u0103, care creeaz\u0103 dificult\u0103\u021bi \u00een anumite condi\u021bii ale solu\u021biei hiperconvergente, unde resursele sunt necesare \u0219i pentru sarcini de calcul \u00een medii virtuale. <\/p>\n<p>Exist\u0103 de asemenea anumite cifre de performan\u021b\u0103 ale Gluster care pot fi ob\u021binute \u00een anumite condi\u021bii, limit\u00e2ndu-se \u00een <noindex><a rel=\"nofollow\" href=\"http:\/\/moo.nac.uci.edu\/~hjm\/Performance_in_a_Gluster_Systemv6F.pdf\"> rezilien\u021b\u0103.<\/a><\/noindex><\/p>\n<h3>Ceph<\/h3>\n<p>\nAcum s\u0103 analiz\u0103m Ceph din descrierile arhitecturii pe care am reu\u0219it <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/313644\/\">s\u0103 le g\u0103sesc.<\/a><\/noindex> De asemenea, exist\u0103 o compara\u021bie \u00eentre <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/company\/croccloudteam\/blog\/430474\/\">Glusterfs \u0219i Ceph<\/a><\/noindex>, unde se poate \u00een\u021belege imediat c\u0103 Ceph este recomandat s\u0103 fie desf\u0103\u0219urat pe servere separate, deoarece serviciile sale necesit\u0103 toate resursele hardware \u00een condi\u021bii de sarcin\u0103. <\/p>\n<p>Arhitectur\u0103 <noindex><a rel=\"nofollow\" href=\"https:\/\/docs.ceph.com\/docs\/mimic\/architecture\/\">Ceph<\/a><\/noindex> mai complicat dec\u00e2t Gluster \u0219i exist\u0103 servicii precum serviciile de metadate, dar \u00eentregul set de componente este destul de complex \u0219i nu foarte flexibil pentru utilizarea sa \u00eentr-o solu\u021bie de virtualizare. Datele sunt organizate \u00een blocuri, ceea ce pare mai performant, dar exist\u0103, \u00een ierarhia tuturor serviciilor (componentelor), pierderi \u0219i laten\u021b\u0103 \u00een condi\u021bii de \u00eenc\u0103rcare \u0219i de urgen\u021b\u0103, de exemplu urm\u0103toarea <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/company\/oleg-bunin\/blog\/431536\/\">articolul.<\/a><\/noindex><\/p>\n<p>Din descrierea arhitecturii, inima este CRUSH, care determin\u0103 locul de stocare a datelor. Urm\u0103toarea este PG \u2013 o abstrac\u021bie destul de complex\u0103 (grup logic) de \u00een\u021beles. PG-urile sunt necesare pentru a face CRUSH mai eficient. Scopul principal al PG-urilor este gruparea obiectelor pentru a reduce consumul de resurse, a \u00eembun\u0103t\u0103\u021bi performan\u021ba \u0219i scalabilitatea. Addrisi obiectele direct, individual, f\u0103r\u0103 a le grupa \u00een PG ar fi foarte costisitor. OSD este serviciul pentru fiecare disc individual.<\/p>\n<p><img decoding=\"async\" alt=\"O compara\u021bie scurt\u0103 a arhitecturii SDS sau c\u0103utarea unei platforme de stocare potrivite (GlusterVsCephVsVirtuozzoStorage)\" src=\"\/wp-content\/uploads\/2020\/02\/b7d59066a01db9bbb55bbd87c6e5053a.png\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<br \/>\n<img decoding=\"async\" alt=\"O compara\u021bie scurt\u0103 a arhitecturii SDS sau c\u0103utarea unei platforme de stocare potrivite (GlusterVsCephVsVirtuozzoStorage)\" src=\"\/wp-content\/uploads\/2020\/02\/a86622bc051ab820d42a8aa08a3fc759.png\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<br \/>\nUn cluster poate avea unul sau mai multe pool-uri de date cu scopuri diferite \u0219i cu set\u0103ri diferite. Pool-urile sunt \u00eemp\u0103r\u021bite \u00een grupuri de plasare. Grupurile de plasare stocheaz\u0103 obiectele la care clien\u021bii se adreseaz\u0103. Aceast\u0103 nivel logic se \u00eencheie \u0219i \u00eencepe cel fizic, deoarece fiecare grup de plasare este asociat cu un disc principal \u0219i mai multe discuri-replici (c\u00e2te exact depinde de factorul de replicare al pool-ului). Cu alte cuvinte, la nivel logic, un obiect este stocat \u00eentr-un anumit grup de plasare, iar la nivel fizic \u2013 pe discurile asociate. \u00cen acela\u0219i timp, discurile pot fi localizate fizic pe noduri diferite sau chiar \u00een centre de date diferite.<\/p>\n<p>\u00cen acest sistem, grupurile de plasare par a fi un nivel necesar pentru flexibilitatea \u00eentregii solu\u021bii, dar \u00een acela\u0219i timp \u0219i un veriga suplimentar\u0103 \u00een aceast\u0103 lan\u021b, ceea ce inevitabil duce la g\u00e2nduri despre pierderi de performan\u021b\u0103. De exemplu, atunci c\u00e2nd scrie date, sistemul trebuie s\u0103 le \u00eempart\u0103 \u00een aceste grupuri \u0219i apoi la nivel fizic pe discul principal \u0219i pe discurile pentru replicare. Asta \u00eenseamn\u0103 c\u0103 func\u021bia hash func\u021bioneaz\u0103 la c\u0103utarea \u0219i inserarea unui obiect, dar exist\u0103 un efect secundar \u2013 costuri foarte mari \u0219i limit\u0103ri pe reconstruirea hash-ului (c\u00e2nd se adaug\u0103 sau se elimin\u0103 un disc). O alt\u0103 problem\u0103 a hash-ului este plasarea strict\u0103 a datelor, care nu poate fi schimbat\u0103. Deci, dac\u0103 un disc se confrunt\u0103 cu o sarcin\u0103 crescut\u0103, sistemul nu are posibilitatea de a nu scrie pe el (aleg\u00e2nd un alt disc), func\u021bia hash oblig\u0103 plasarea datelor conform unei reguli, indiferent de c\u00e2t de r\u0103u este pentru disc, astfel \u00eenc\u00e2t Ceph consum\u0103 mult\u0103 memorie la reconstruirea PG \u00een cazul de auto-repair sau extinderea stoc\u0103rii. Concluzia este c\u0103 Ceph func\u021bioneaz\u0103 bine (chiar dac\u0103 lent), dar doar c\u00e2nd nu sunt scal\u0103ri, situa\u021bii de urgen\u021b\u0103 sau actualiz\u0103ri.<\/p>\n<p>Exist\u0103, desigur, op\u021biuni pentru \u00eembun\u0103t\u0103\u021birea performan\u021bei prin intermediul caching-ului \u0219i tiering-ului de cache, dar pentru asta este necesar un hardware bun \u0219i tot vor exista pierderi. Dar, \u00een ansamblu, Ceph pare mai atr\u0103g\u0103tor dec\u00e2t Gluster pentru produc\u021bie. De asemenea, utilizarea acestor produse necesit\u0103 un factor important \u2013 un nivel ridicat de competen\u021be, experien\u021b\u0103 \u0219i profesionalism, cu un accent puternic pe Linux, deoarece este foarte important s\u0103 fie implementat, configurat \u0219i \u00eentre\u021binut corect, ceea ce impune \u0219i mai multe responsabilit\u0103\u021bi \u0219i presiuni asupra administratorului. <\/p>\n<h3>Vstorage<\/h3>\n<p>\nArhitectura devine \u0219i mai interesant\u0103 \u00een cazul <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/company\/virtuozzo\/blog\/334724\/\">Virtuozzo storage (Vstorage)<\/a><\/noindex>, care poate fi utilizat\u0103 \u00eempreun\u0103 cu hipervizorul pe acelea\u0219i noduri, pe acela\u0219i <noindex><a rel=\"nofollow\" href=\"http:\/\/samag.ru\/archive\/article\/3794\">hardware<\/a><\/noindex>, dar este foarte important s\u0103 fie configurat corect pentru a ob\u021bine o performan\u021b\u0103 bun\u0103. A\u0219adar, implement\u00e2nd un astfel de produs din cutie pe o configura\u021bie oarecare, f\u0103r\u0103 a \u021bine cont de recomand\u0103rile conform arhitecturii, va fi foarte u\u0219or, dar nu eficient.<\/p>\n<p>Ce poate coexista cu serviciile hipervizorului kvm-qemu sunt doar c\u00e2teva servicii, unde se g\u0103se\u0219te o ierarhie optimizat\u0103 compact\u0103 de componente: serviciul clientului montat prin FUSE (modificat, nu open source), serviciul de metadate MDS (Metadata service), serviciul de blocuri de date Chunk service, care la nivel fizic echivaleaz\u0103 cu un singur disc \u0219i cam at\u00e2t. Desigur, pentru vitez\u0103, este optim s\u0103 folose\u0219ti o schem\u0103 de toleran\u021b\u0103 la e\u0219ec \u00een dou\u0103 replici, dar dac\u0103 apelez la caching \u0219i jurnale pe discuri SSD, atunci codificarea rezistent\u0103 la erori (erase coding sau raid6) poate fi accelerat\u0103 bine pe o schem\u0103 hibrid\u0103 sau chiar mai bine pe all flash. Cu EC (erase coding) exist\u0103 un dezavantaj: la modificarea unui bloc de date, este necesar s\u0103 se recalculeze sumele de paritate. Pentru a evita pierderile \u00een aceast\u0103 opera\u021biune, Ceph scrie \u00een EC \u00een mod \u00eent\u00e2rziat \u0219i pot ap\u0103rea probleme de performan\u021b\u0103 la anumite cereri, c\u00e2nd, de exemplu, este nevoie s\u0103 se calculeze toate blocurile, iar \u00een cazul cu Virtuozzo Storage, scrierea blocurilor modificate se face folosind abordarea \u201esistem de fi\u0219iere structurat pe jurnal\u201d, ceea ce minimizeaz\u0103 costurile de calcul al parit\u0103\u021bii. Pentru a aproxima op\u021biunile cu accelerarea muncii cu EC \u0219i f\u0103r\u0103, exist\u0103 <noindex><a rel=\"nofollow\" href=\"https:\/\/www.virtuozzo.com\/products\/virtuozzo-storage\/storage-calculator\">un calculator.<\/a><\/noindex> \u2013 cifrele pot fi approximate, depind de coeficientul de precizie al produc\u0103torului echipamentului, dar rezultatul calculelor ajut\u0103 foarte bine la planificarea configura\u021biei.<\/p>\n<p>O schem\u0103 simpl\u0103 a componentelor de stocare nu \u00eenseamn\u0103 c\u0103 aceste componente nu consum\u0103 <noindex><a rel=\"nofollow\" href=\"https:\/\/docs.virtuozzo.com\/virtuozzo_7_installation_guide\/preparing-for-installation\/planning-storage-cli.html#hardware-requirements\">resurse hardware,<\/a><\/noindex> dar dac\u0103 socote\u0219ti toate cheltuielile din timp, atunci po\u021bi conta pe colaborarea de l\u00e2ng\u0103 hipervizor. <br \/>\nExist\u0103 o schem\u0103 de compara\u021bie a consumului de resurse hardware de c\u0103tre serviciile Ceph \u0219i Virtuozzo storage.<\/p>\n<p><img decoding=\"async\" alt=\"O compara\u021bie scurt\u0103 a arhitecturii SDS sau c\u0103utarea unei platforme de stocare potrivite (GlusterVsCephVsVirtuozzoStorage)\" src=\"\/wp-content\/uploads\/2020\/02\/27de0ab0cc61c2fab6d741bc3105b6cc.png\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<br \/>\nDac\u0103 anterior puteai compara Gluster \u0219i Ceph pe baza articolelor vechi, folosind cele mai importante pasaje din ele, cu Virtuozzo este mai complicat. Articole despre acest produs nu sunt foarte multe \u0219i informa\u021biile pot fi extrase doar din documenta\u021bia pe <noindex><a rel=\"nofollow\" href=\"https:\/\/docs.virtuozzo.com\/master\/index.html\">englez\u0103<\/a><\/noindex> sau \u00een limba rus\u0103, av\u00e2nd \u00een vedere c\u0103 Vstorage este utilizat \u00een unele solu\u021bii hiperconvergente \u00een companii precum <noindex><a rel=\"nofollow\" href=\"http:\/\/rosplatforma.ru\/#downloads\">Rospplatforma<\/a><\/noindex> \u0219i Acronis.<\/p>\n<p>Voi \u00eencerca s\u0103 ajut cu descrierea acestei arhitecturi, de aceea textul va fi pu\u021bin mai lung, dar pentru a \u00een\u021belege documenta\u021bia este nevoie de mult timp, iar documenta\u021bia existent\u0103 poate fi folosit\u0103 doar ca un ghid prin revizuirea cuprinsului sau c\u0103utarea dup\u0103 cuv\u00e2nt cheie. <\/p>\n<p>S\u0103 consider\u0103m procesul de scriere \u00eentr-o configura\u021bie hibrid\u0103 de hardware cu componentele descrise mai sus: scrierea \u00eencepe pe acel nod de la care a fost ini\u021biat\u0103 de client (serviciul punctului de montare FUSE), dar componenta master a serviciului de metadate (MDS) va direc\u021biona desigur clientul direct c\u0103tre serviciul de chunk necesar (serviciul de stocare a blocurilor CS), ceea ce \u00eenseamn\u0103 c\u0103 MDS nu particip\u0103 la procesul de scriere, ci pur \u0219i simplu direc\u021bioneaz\u0103 c\u0103tre serviciul de chunk necesar. \u00cen general, se poate aduce o analogie a scrierii cu turnarea apei \u00een butoaie. Fiecare butoi reprezint\u0103 un bloc de date de 256MB.<\/p>\n<p><img decoding=\"async\" alt=\"O compara\u021bie scurt\u0103 a arhitecturii SDS sau c\u0103utarea unei platforme de stocare potrivite (GlusterVsCephVsVirtuozzoStorage)\" src=\"\/wp-content\/uploads\/2020\/02\/8f10333f1573226fbb50a40377eafc93.png\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<br \/>\nA\u0219adar, un disc reprezint\u0103 o anumit\u0103 cantitate de astfel de butoaie, adic\u0103 volumul discului \u00eemp\u0103r\u021bit la 256MB. Fiecare copie se toarn\u0103 pe un nod, iar a doua aproape simultan pe un alt nod etc. Dac\u0103 avem trei replici \u0219i avem discuri SSD pentru cache (pentru citire \u0219i jurnalele de scriere), confirmarea scrierii va avea loc dup\u0103 scrierea jurnalului \u00een SSD, iar desc\u0103rcarea paralel\u0103 din SSD va continua pe HDD, \u00eentr-un fel de mod \u00een fundal. \u00cen cazul celor trei replici, confirmarea scrierii va fi dup\u0103 confirmarea din partea SSD-ului de pe al treilea nod. Ar putea p\u0103rea c\u0103 suma vitezei de scriere a celor trei SSD-uri poate fi \u00eemp\u0103r\u021bit\u0103 la trei \u0219i vom ob\u021bine viteza de scriere a unei replici, dar scrierea copiilor se desf\u0103\u0219oar\u0103 \u00een paralel, iar latenta re\u021belei este de obicei mai mare dec\u00e2t cea a SSD-urilor, iar, \u00een esen\u021b\u0103, performan\u021ba scrierii va depinde de re\u021bea. Din aceast\u0103 cauz\u0103, pentru a observa IOPS reale este necesar s\u0103 se \u00eencarce corect \u00eentreg Vstorage dup\u0103 <noindex><a rel=\"nofollow\" href=\"https:\/\/docs.virtuozzo.com\/pdf\/virtuozzo_infrastructure_platform_3_benchmarking_guide.pdf\">metodica<\/a><\/noindex>, adic\u0103 s\u0103 se testeze o \u00eenc\u0103rcare real\u0103, nu doar memoria \u0219i cache-ul, unde trebuie s\u0103 se ia \u00een considerare dimensiunea corect\u0103 a blocului de date, num\u0103rul de fire \u0219i a\u0219a mai departe. <\/p>\n<p>Jurnalul de \u00eenregistrare men\u021bionat mai sus pe SSD func\u021bioneaz\u0103 astfel: de \u00eendat\u0103 ce datele ajung \u00een el, sunt citite imediat de serviciu \u0219i scrise pe HDD. Exist\u0103 mai multe servicii de metadate (MDS) pe cluster, iar num\u0103rul acestora este determinat de cvorum, care func\u021bioneaz\u0103 conform algoritmului Paxos. Din perspectiva clientului, punctul de montare FUSE este un folder al stoc\u0103rii clusterului, care este vizibil simultan pentru toate nodurile clusterului, fiecare nod av\u00e2nd un client montat pe acest principiu, astfel \u00eenc\u00e2t fiecare nod are acces la aceast\u0103 stocare.<\/p>\n<p>Pentru performan\u021ba oric\u0103ruia dintre abord\u0103rile men\u021bionate mai sus, este foarte important, \u00een etapa de planificare \u0219i desf\u0103\u0219urare, s\u0103 se configureze corect re\u021beaua, unde va exista un echilibru datorat agreg\u0103rii \u0219i o l\u0103\u021bime de band\u0103 a canalului de re\u021bea bine aleas\u0103. \u00cen agregare, este important s\u0103 se aleag\u0103 corect modul de hashing \u0219i dimensiunile cadrelor. De asemenea, exist\u0103 o diferen\u021b\u0103 semnificativ\u0103 fa\u021b\u0103 de SDS-urile men\u021bionate anterior, aceasta fiind fuse cu tehnologia fast path \u00een Virtuozzo Storage. Aceasta, pe l\u00e2ng\u0103 fuse-urile modernizate, spre deosebire de celelalte solu\u021bii open source, adaug\u0103 semnificativ IOPS-uri \u0219i permite s\u0103 nu te limitezi la scalarea orizontal\u0103 sau vertical\u0103. \u00cen general, comparativ cu arhitecturile men\u021bionate mai sus, aceasta pare a fi mai puternic\u0103, dar pentru acest avantaj este necesar s\u0103 achizi\u021bionezi licen\u021be, spre deosebire de Ceph \u0219i Gluster. <\/p>\n<p>\u00cen concluzie, putem sublinia clasamentul top trei: prima pozi\u021bie \u00een ceea ce prive\u0219te performan\u021ba \u0219i fiabilitatea arhitecturii este ocupat\u0103 de Virtuozzo Storage, a doua de Ceph \u0219i a treia de Gluster. <\/p>\n<p>Criteriile pentru alegerea Virtuozzo Storage sunt: un set optim de componente arhitecturale, fuse modernizat sub acest principiu cu fast path, un set flexibil de configura\u021bie hardware, un consum mai mic de resurse \u0219i posibilitatea de a fi utilizat \u00eempreun\u0103 cu compute (calcul\u0103ri\/virtualizare), deci este complet potrivit pentru o solu\u021bie hiperconvergent\u0103, din cadrul c\u0103reia face parte. A doua pozi\u021bie este ocupat\u0103 de Ceph, deoarece este o arhitectur\u0103 mai performant\u0103 dec\u00e2t Gluster, datorit\u0103 oper\u0103rii pe blocuri, precum \u0219i a scenariilor mai flexibile \u0219i a posibilit\u0103\u021bii de a func\u021biona \u00een clustere de dimensiuni mai mari.<\/p>\n<p>\u00cen planuri exist\u0103 dorin\u021ba de a scrie o compara\u021bie \u00eentre vSAN, Space Direct Storage, Vstorage \u0219i Nutanix Storage, testarea Vstorage pe echipamente HPE, Huawei, precum \u0219i scenarii de integrare a Vstorage cu sisteme externe de stocare hardware, a\u0219a c\u0103, dac\u0103 articolul v-a pl\u0103cut, ar fi bine s\u0103 primim feedback de la voi, care ar putea \u00eent\u0103ri motiva\u021bia pentru noi articole \u021bin\u00e2nd cont de observa\u021biile \u0219i dorin\u021bele voastre.<br \/>\n<br \/>Sursa: <a content=\"nofollow\" rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/486392\/\">habr.com<\/a> <\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"<p>\u0414\u0430\u043d\u043d\u0430\u044f \u0441\u0442\u0430\u0442\u044c\u044f \u043d\u0430\u043f\u0438\u0441\u0430\u043d\u0430 \u0434\u043b\u044f \u0442\u043e\u0433\u043e, \u0447\u0442\u043e\u0431\u044b \u043f\u043e\u043c\u043e\u0447\u044c \u0432\u044b\u0431\u0440\u0430\u0442\u044c \u0434\u043b\u044f \u0441\u0435\u0431\u044f \u043f\u043e\u0434\u0445\u043e\u0434\u044f\u0449\u0435\u0435 \u0440\u0435\u0448\u0435\u043d\u0438\u0435 \u0438 \u043f\u043e\u043d\u044f\u0442\u044c \u043e\u0442\u043b\u0438\u0447\u0438\u044f \u043c\u0435\u0436\u0434\u0443 \u0442\u0430\u043a\u0438\u043c\u0438 SDS \u043a\u0430\u043a Gluster, Ceph \u0438 Vstorage (Virtuozzo). \u0412 \u0442\u0435\u043a\u0441\u0442\u0435 \u0438\u0441\u043f\u043e\u043b\u044c\u0437\u0443\u044e\u0442\u0441\u044f \u0441\u0441\u044b\u043b\u043a\u0438 \u043d\u0430 \u0441\u0442\u0430\u0442\u044c\u0438 \u0441 \u0431\u043e\u043b\u0435\u0435 \u0434\u0435\u0442\u0430\u043b\u044c\u043d\u044b\u043c \u0440\u0430\u0441\u043a\u0440\u044b\u0442\u0438\u0435\u043c \u0442\u0435\u0445 \u0438\u043b\u0438 \u0438\u043d\u044b\u0445 \u043f\u0440\u043e\u0431\u043b\u0435\u043c, \u043f\u043e\u044d\u0442\u043e\u043c\u0443 \u043e\u043f\u0438\u0441\u0430\u043d\u0438\u044f \u0431\u0443\u0434\u0443\u0442 \u043c\u0430\u043a\u0441\u0438\u043c\u0430\u043b\u044c\u043d\u043e \u043a\u0440\u0430\u0442\u043a\u0438\u043c\u0438 \u0441 \u0438\u0441\u043f\u043e\u043b\u044c\u0437\u043e\u0432\u0430\u043d\u0438\u0435\u043c \u043a\u043b\u044e\u0447\u0435\u0432\u044b\u0445 \u043c\u043e\u043c\u0435\u043d\u0442\u043e\u0432 \u0431\u0435\u0437 \u043b\u0438\u0448\u043d\u0435\u0439 \u0432\u043e\u0434\u044b \u0438 \u0432\u0432\u043e\u0434\u043d\u043e\u0439 \u0438\u043d\u0444\u043e\u0440\u043c\u0430\u0446\u0438\u0438, \u043a\u043e\u0442\u043e\u0440\u0443\u044e \u0432\u044b [&hellip;]<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"author":1,"featured_media":40730,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[],"tags":[],"class_list":["post-40729","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry"],"aioseo_notices":[],"aioseo_head":"\n\t\t<!-- All in One SEO 5.0.2.1 - aioseo.com -->\n\t<meta name=\"description\" content=\"\u0414\u0430\u043d\u043d\u0430\u044f \u0441\u0442\u0430\u0442\u044c\u044f \u043d\u0430\u043f\u0438\u0441\u0430\u043d\u0430 \u0434\u043b\u044f \u0442\u043e\u0433\u043e, \u0447\u0442\u043e\u0431\u044b \u043f\u043e\u043c\u043e\u0447\u044c \u0432\u044b\u0431\u0440\u0430\u0442\u044c \u0434\u043b\u044f \u0441\u0435\u0431\u044f \u043f\u043e\u0434\u0445\u043e\u0434\u044f\u0449\u0435\u0435 \u0440\u0435\u0448\u0435\u043d\u0438\u0435 \u0438 \u043f\u043e\u043d\u044f\u0442\u044c \u043e\u0442\u043b\u0438\u0447\u0438\u044f \u043c\u0435\u0436\u0434\u0443 \u0442\u0430\u043a\u0438\u043c\u0438 SDS \u043a\u0430\u043a Gluster, Ceph \u0438 Vstorage (Virtuozzo). \u0412 \u0442\u0435\u043a\u0441\u0442\u0435 \u0438\u0441\u043f\u043e\u043b\u044c\u0437\u0443\u044e\u0442\u0441\u044f \u0441\u0441\u044b\u043b\u043a\u0438 \u043d\u0430 \u0441\u0442\u0430\u0442\u044c\u0438 \u0441 \u0431\u043e\u043b\u0435\u0435.\" \/>\n\t<meta name=\"robots\" content=\"max-image-preview:large\" \/>\n\t<meta name=\"author\" content=\"Yuri Gagarin\"\/>\n\t<link rel=\"canonical\" href=\"https:\/\/prohoster.info\/ro\/blog\/kratkoe-sravnenie-arhitektury-sds-ili-poisk-podhodyashhej-platformy-hraneniya-glustervscephvsvirtuozzostorage\" \/>\n\t<meta name=\"generator\" content=\"All in One SEO (AIOSEO) 5.0.2.1\" \/>\n\t\t<meta property=\"og:locale\" content=\"ro_RO\" \/>\n\t\t<meta property=\"og:site_name\" content=\"ProHoster | \u041a\u0443\u043f\u0438\u0442\u044c \u043d\u0430\u0434\u0435\u0436\u043d\u044b\u0439 \u0445\u043e\u0441\u0442\u0438\u043d\u0433 \u0434\u043b\u044f \u0441\u0430\u0439\u0442\u043e\u0432 \u0441 \u0437\u0430\u0449\u0438\u0442\u043e\u0439 \u043e\u0442 DDoS, VPS VDS \u0441\u0435\u0440\u0432\u0435\u0440\u044b\" \/>\n\t\t<meta property=\"og:type\" content=\"article\" \/>\n\t\t<meta property=\"og:title\" content=\"\ud83e\udd47\u041a\u0440\u0430\u0442\u043a\u043e\u0435 \u0441\u0440\u0430\u0432\u043d\u0435\u043d\u0438\u0435 \u0430\u0440\u0445\u0438\u0442\u0435\u043a\u0442\u0443\u0440\u044b SDS \u0438\u043b\u0438 \u043f\u043e\u0438\u0441\u043a \u043f\u043e\u0434\u0445\u043e\u0434\u044f\u0449\u0435\u0439 \u043f\u043b\u0430\u0442\u0444\u043e\u0440\u043c\u044b \u0445\u0440\u0430\u043d\u0435\u043d\u0438\u044f (GlusterVsCephVsVirtuozzoStorage) | ProHoster\" \/>\n\t\t<meta property=\"og:description\" content=\"\u0414\u0430\u043d\u043d\u0430\u044f \u0441\u0442\u0430\u0442\u044c\u044f \u043d\u0430\u043f\u0438\u0441\u0430\u043d\u0430 \u0434\u043b\u044f \u0442\u043e\u0433\u043e, \u0447\u0442\u043e\u0431\u044b \u043f\u043e\u043c\u043e\u0447\u044c \u0432\u044b\u0431\u0440\u0430\u0442\u044c \u0434\u043b\u044f \u0441\u0435\u0431\u044f \u043f\u043e\u0434\u0445\u043e\u0434\u044f\u0449\u0435\u0435 \u0440\u0435\u0448\u0435\u043d\u0438\u0435 \u0438 \u043f\u043e\u043d\u044f\u0442\u044c \u043e\u0442\u043b\u0438\u0447\u0438\u044f \u043c\u0435\u0436\u0434\u0443 \u0442\u0430\u043a\u0438\u043c\u0438 SDS \u043a\u0430\u043a Gluster, Ceph \u0438 Vstorage (Virtuozzo). \u0412 \u0442\u0435\u043a\u0441\u0442\u0435 \u0438\u0441\u043f\u043e\u043b\u044c\u0437\u0443\u044e\u0442\u0441\u044f \u0441\u0441\u044b\u043b\u043a\u0438 \u043d\u0430 \u0441\u0442\u0430\u0442\u044c\u0438 \u0441 \u0431\u043e\u043b\u0435\u0435.\" \/>\n\t\t<meta property=\"og:url\" content=\"https:\/\/prohoster.info\/ro\/blog\/kratkoe-sravnenie-arhitektury-sds-ili-poisk-podhodyashhej-platformy-hraneniya-glustervscephvsvirtuozzostorage\" \/>\n\t\t<meta property=\"og:image\" content=\"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg\" \/>\n\t\t<meta property=\"og:image:secure_url\" content=\"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg\" \/>\n\t\t<meta property=\"og:image:width\" content=\"350\" \/>\n\t\t<meta property=\"og:image:height\" content=\"350\" \/>\n\t\t<meta property=\"article:published_time\" content=\"2020-02-03T11:42:16+00:00\" \/>\n\t\t<meta property=\"article:modified_time\" content=\"2020-02-03T11:42:16+00:00\" \/>\n\t\t<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/prohoster\" \/>\n\t\t<meta property=\"article:author\" content=\"https:\/\/www.facebook.com\/prohoster\" \/>\n\t\t<!-- All in One SEO -->\n\n","aioseo_head_json":{"title":"\ud83e\udd47Compara\u021bie scurt\u0103 a arhitecturii SDS sau c\u0103utarea unei platforme de stocare potrivite (GlusterVsCephVsVirtuozzoStorage) | ProHoster","description":"Acest articol a fost scris pentru a ajuta la alegerea unei solu\u021bii potrivite \u0219i pentru a \u00een\u021belege diferen\u021bele dintre SDS precum Gluster, Ceph \u0219i Vstorage (Virtuozzo). \u00cen text sunt folosite linkuri c\u0103tre articole suplimentare.","canonical_url":"https:\/\/prohoster.info\/ro\/blog\/kratkoe-sravnenie-arhitektury-sds-ili-poisk-podhodyashhej-platformy-hraneniya-glustervscephvsvirtuozzostorage","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"ro_RO","og:site_name":"ProHoster | \u041a\u0443\u043f\u0438\u0442\u044c \u043d\u0430\u0434\u0435\u0436\u043d\u044b\u0439 \u0445\u043e\u0441\u0442\u0438\u043d\u0433 \u0434\u043b\u044f \u0441\u0430\u0439\u0442\u043e\u0432 \u0441 \u0437\u0430\u0449\u0438\u0442\u043e\u0439 \u043e\u0442 DDoS, VPS VDS \u0441\u0435\u0440\u0432\u0435\u0440\u044b","og:type":"article","og:title":"\ud83e\udd47\u041a\u0440\u0430\u0442\u043a\u043e\u0435 \u0441\u0440\u0430\u0432\u043d\u0435\u043d\u0438\u0435 \u0430\u0440\u0445\u0438\u0442\u0435\u043a\u0442\u0443\u0440\u044b SDS \u0438\u043b\u0438 \u043f\u043e\u0438\u0441\u043a \u043f\u043e\u0434\u0445\u043e\u0434\u044f\u0449\u0435\u0439 \u043f\u043b\u0430\u0442\u0444\u043e\u0440\u043c\u044b \u0445\u0440\u0430\u043d\u0435\u043d\u0438\u044f (GlusterVsCephVsVirtuozzoStorage) | ProHoster","og:description":"\u0414\u0430\u043d\u043d\u0430\u044f \u0441\u0442\u0430\u0442\u044c\u044f \u043d\u0430\u043f\u0438\u0441\u0430\u043d\u0430 \u0434\u043b\u044f \u0442\u043e\u0433\u043e, \u0447\u0442\u043e\u0431\u044b \u043f\u043e\u043c\u043e\u0447\u044c \u0432\u044b\u0431\u0440\u0430\u0442\u044c \u0434\u043b\u044f \u0441\u0435\u0431\u044f \u043f\u043e\u0434\u0445\u043e\u0434\u044f\u0449\u0435\u0435 \u0440\u0435\u0448\u0435\u043d\u0438\u0435 \u0438 \u043f\u043e\u043d\u044f\u0442\u044c \u043e\u0442\u043b\u0438\u0447\u0438\u044f \u043c\u0435\u0436\u0434\u0443 \u0442\u0430\u043a\u0438\u043c\u0438 SDS \u043a\u0430\u043a Gluster, Ceph \u0438 Vstorage (Virtuozzo). \u0412 \u0442\u0435\u043a\u0441\u0442\u0435 \u0438\u0441\u043f\u043e\u043b\u044c\u0437\u0443\u044e\u0442\u0441\u044f \u0441\u0441\u044b\u043b\u043a\u0438 \u043d\u0430 \u0441\u0442\u0430\u0442\u044c\u0438 \u0441 \u0431\u043e\u043b\u0435\u0435.","og:url":"https:\/\/prohoster.info\/ro\/blog\/kratkoe-sravnenie-arhitektury-sds-ili-poisk-podhodyashhej-platformy-hraneniya-glustervscephvsvirtuozzostorage","og:image":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:secure_url":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:width":350,"og:image:height":350,"article:published_time":"2020-02-03T11:42:16+00:00","article:modified_time":"2020-02-03T11:42:16+00:00","article:publisher":"https:\/\/www.facebook.com\/prohoster","article:author":"https:\/\/www.facebook.com\/prohoster"},"aioseo_meta_data":{"post_id":"40729","title":null,"description":null,"keywords":null,"keyphrases":null,"primary_term":null,"canonical_url":null,"og_title":null,"og_description":null,"og_object_type":"default","og_image_type":"default","og_image_url":null,"og_image_width":null,"og_image_height":null,"og_image_custom_url":null,"og_image_custom_fields":null,"og_video":null,"og_custom_url":null,"og_article_section":null,"og_article_tags":null,"twitter_use_og":false,"twitter_card":"default","twitter_image_type":"default","twitter_image_url":null,"twitter_image_custom_url":null,"twitter_image_custom_fields":null,"twitter_title":null,"twitter_description":null,"schema":{"blockGraphs":[],"customGraphs":[],"default":{"data":{"Article":[],"Course":[],"Dataset":[],"FAQPage":[],"Movie":[],"Person":[],"Product":[],"ProductReview":[],"Car":[],"Recipe":[],"Service":[],"SoftwareApplication":[],"WebPage":[]},"graphName":"","isEnabled":true},"graphs":[]},"schema_type":null,"schema_type_options":null,"pillar_content":false,"robots_default":true,"robots_noindex":false,"robots_noarchive":false,"robots_nosnippet":false,"robots_nofollow":false,"robots_noimageindex":false,"robots_noodp":false,"robots_notranslate":false,"robots_max_snippet":null,"robots_max_videopreview":null,"robots_max_imagepreview":"large","priority":null,"frequency":null,"local_seo":null,"seo_analyzer_scan_date":null,"breadcrumb_settings":null,"limit_modified_date":false,"reviewed_by":null,"ai":null,"created":"2021-03-01 00:32:40","updated":"2022-09-29 09:23:33","focus_keyword":null,"additional_keywords":null,"truseo_locale":null},"gt_translate_keys":[{"key":"link","format":"url"}],"_links":{"self":[{"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/posts\/40729","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/comments?post=40729"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/posts\/40729\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/media\/40730"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/media?parent=40729"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/categories?post=40729"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/tags?post=40729"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}