{"id":42117,"date":"2019-03-18T00:00:00","date_gmt":"2019-03-17T21:00:00","guid":{"rendered":"https:\/\/prohoster.info\/blog\/blog_prohoster\/situatsiya-virtualnye-gpu-ne-ustupayut-po-proizvoditelnosti-zheleznym-resheniyam"},"modified":"2020-02-18T13:44:55","modified_gmt":"2020-02-18T10:44:55","slug":"situatsiya-virtualnye-gpu-ne-ustupayut-po-proizvoditelnosti-zheleznym-resheniyam","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/administrirovanie\/situatsiya-virtualnye-gpu-ne-ustupayut-po-proizvoditelnosti-zheleznym-resheniyam","title":{"rendered":"Situata: GPU virtuale nuk i japin nga performanca zgjidhjeve fizike.","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

N\u00eb shkurt u mbajt nj\u00eb konferenc\u00eb n\u00eb Stanford, e dedikuar kompjuterave me performanc\u00eb t\u00eb lart\u00eb (HPC). P\u00ebrfaq\u00ebsuesit e VMware informuan se gjat\u00eb p\u00ebrdorimit t\u00eb GPU-ve, sistemi i bazuar n\u00eb nj\u00eb hipervizor t\u00eb modifikuar ESXi, renditet nj\u00eblloj si zgjidhjet bare metal n\u00eb terma shpejt\u00ebsie.<\/p>\n

Po flasim p\u00ebr teknologjit\u00eb q\u00eb e b\u00ebn\u00eb k\u00ebt\u00eb t\u00eb mundur.<\/p>\n

\"Situata:<\/a><\/noindex><\/a><\/noindex>
\n\/ \u0444\u043e\u0442\u043e Victorgrigas<\/a><\/noindex> CC BY-SA<\/a><\/noindex><\/i><\/p>\n

Problemi i performanc\u00ebs<\/h2>\n

\nSipas analist\u00ebve, rreth 70% e ngarkesave t\u00eb pun\u00ebs n\u00eb qendrat e t\u00eb dh\u00ebnave jan\u00eb virtualizuar<\/a><\/noindex>. Megjithat\u00eb, 30% e mbetur ende punon n\u00eb bare metal pa hipervizor\u00eb. K\u00ebto 30% kryesisht p\u00ebrb\u00ebhen nga aplikacione me ngarkes\u00eb t\u00eb lart\u00eb, t\u00eb lidhura, p\u00ebr shembull, me trajnimin e rrjeteve neuronale, dhe q\u00eb p\u00ebrdorin GPU-t\u00eb.<\/p>\n

Ekspert\u00ebt e shpjegojn\u00eb k\u00ebt\u00eb trend me faktin se hipervizori si nj\u00eb shtres\u00eb nd\u00ebrmjet\u00ebse abstraksioni mund t\u00eb ndikoj\u00eb n\u00eb performanc\u00ebn e t\u00eb gjith\u00eb sistemit. N\u00eb studimet e pes\u00eb viteve m\u00eb par\u00eb mund t\u00eb gjenden t\u00eb dh\u00ebna<\/a><\/noindex> p\u00ebr nj\u00eb ulje t\u00eb shpejt\u00ebsis\u00eb s\u00eb pun\u00ebs prej 10%. Prandaj, kompanit\u00eb dhe operator\u00ebt e qendrave t\u00eb t\u00eb dh\u00ebnave nuk nxitohen t\u00eb transfertojn\u00eb ngarkesat HPC n\u00eb ambientin virtual.<\/p>\n

Por teknologjit\u00eb e virtualizimit po zhvillohen dhe p\u00ebrmir\u00ebsohen. N\u00eb konferenc\u00ebn e muajit t\u00eb kaluar, VMware tha se hipervizori ESXi nuk ka nj\u00eb nd Wirkung negativ n\u00eb performanc\u00ebn e GPU-s\u00eb. Shpejt\u00ebsia e llogaritjeve mund t\u00eb bjer\u00eb me tre p\u00ebr qind, e cila \u00ebsht\u00eb e krahasueshme me treguesit e bare metal.<\/p>\n

Si funksionon kjo<\/h2>\n

\nP\u00ebr t\u00eb rritur performanc\u00ebn e sistemeve HPC me GPU-t\u00eb, VMware kan\u00eb b\u00ebr\u00eb disa ndryshime n\u00eb funksionimin e hipervizorit. N\u00eb ve\u00e7anti, e kan\u00eb \u00e7liruar at\u00eb nga funksioni vMotion. Ky funksion \u00ebsht\u00eb i nevojsh\u00ebm p\u00ebr balancimin e ngarkes\u00ebs dhe zakonisht transferon makinat virtuale (VM) midis server\u00ebve ose GPU-ve. \u00c7aktivizimi i vMotion ka b\u00ebr\u00eb q\u00eb secila VM tani t\u00eb lidhet me nj\u00eb GPU specifike. Kjo ndihmoi n\u00eb uljen e kostove gjat\u00eb shk\u00ebmbimit t\u00eb t\u00eb dh\u00ebnave.<\/p>\n

Nj\u00eb komponent tjet\u00ebr ky\u00e7 i sistemit \u00ebsht\u00eb teknologjia<\/a><\/noindex> DirectPath I\/O. Kjo lejon q\u00eb drivere e CUDA p\u00ebr llogaritje paralele t\u00eb interagoj\u00eb me makinat virtuale drejtp\u00ebrdrejt, duke kaluar hipervizorin. Kur nj\u00eb GPU ka nevoj\u00eb t\u00eb ekzekutoj\u00eb disa VM nj\u00ebkoh\u00ebsisht, at\u00ebher\u00eb zgjidhet zgjidhja GRID vGPU. Kjo ndan memorie t\u00eb kart\u00ebs n\u00eb disa segmente (por ciklet llogarit\u00ebse nuk ndahen n\u00eb k\u00ebt\u00eb rast).<\/p>\n

Skema e pun\u00ebs s\u00eb dy makinave virtuale n\u00eb k\u00ebt\u00eb rast do t\u00eb duket si vijon:<\/p>\n

\"Situata:<\/p>\n

Rezultatet dhe parashikimet<\/h2>\n

\nKompania kryen teste<\/a><\/noindex> hypervisor, duke modelet nj\u00eb model gjuhe mbi TensorFlow<\/a><\/noindex>. \"D\u00ebmi\" i performanc\u00ebs ishte vet\u00ebm 3-4%, krahasuar me bare metal. Megjithat\u00eb, sistemi fitoi mund\u00ebsin\u00eb p\u00ebr t\u00eb shp\u00ebrndar\u00eb burimet sipas k\u00ebrkes\u00ebs, n\u00eb var\u00ebsi t\u00eb ngarkesave aktuale.<\/p>\n

Giganti IT gjithashtu kryente teste<\/a><\/noindex> me kontejner\u00ebt. Inxhinier\u00ebt e kompanis\u00eb trajnuan rrjetet neuronale p\u00ebr t\u00eb njohur imazhet. N\u00eb k\u00ebt\u00eb proces, burimet e nj\u00eb procesori grafik u shp\u00ebrndan\u00eb mes kat\u00ebr VM-ve kontejnerike. Si rezultat, performanca e makinave individuale u ul me 17% (krahasuar me nj\u00eb VM q\u00eb kishte qasje t\u00eb plot\u00eb n\u00eb burimet GPU). Megjithat\u00eb, numri i imazheve t\u00eb procesuar p\u00ebr sekond\u00eb u rrit<\/a><\/noindex> u trefishua. Pritet q\u00eb sisteme t\u00eb tilla t\u00eb gjejn\u00eb<\/a><\/noindex> p\u00ebrdorim n\u00eb fush\u00ebn e analiz\u00ebs s\u00eb t\u00eb dh\u00ebnave dhe modelimit kompjuterik.<\/p>\n

Mes problemeve t\u00eb mundshme me t\u00eb cilat VMware mund t\u00eb p\u00ebrballet, ekspert\u00ebt dallohet<\/a><\/noindex> n\u00eb nj\u00eb audienc\u00eb mjaft t\u00eb ngusht\u00eb. Aktualisht, numri i kompanive q\u00eb punojn\u00eb me sisteme me performanc\u00eb t\u00eb lart\u00eb \u00ebsht\u00eb i vog\u00ebl. Megjithat\u00eb, sipas Statista v\u00ebrehet<\/a><\/noindex>, deri n\u00eb vitin 2021, 94% e ngarkesave t\u00eb pun\u00ebs bot\u00ebrore do t\u00eb jen\u00eb t\u00eb virtualizuara. Sipas parashikimeve<\/a><\/noindex> analist\u00ebve, vlera e tregut t\u00eb HPC do t\u00eb rritet nga 32 n\u00eb 45 miliard dollar\u00eb n\u00eb periudh\u00ebn nga 2017 deri n\u00eb 2022.<\/p>\n

\"Situata:
\n\/ \u0444\u043e\u0442\u043e Sistemet e ngjashme<\/a><\/noindex> PD<\/i><\/p>\n

Zgjidhje t\u00eb ngjashme<\/h2>\n

\nN\u00eb treg ka disa alternativa, t\u00eb cilat po zhvillohen nga kompani t\u00eb m\u00ebdha IT: AMD dhe Intel.<\/p>\n

Kompania e par\u00eb p\u00ebr virtualizimin e procesor\u00ebve grafik\u00eb ofron<\/noindex> ka qasje mbi baz\u00ebn e SR-IOV (virtualizimi i hyrjes\/daljes me rr\u00ebnj\u00eb t\u00eb vetme). Kjo teknologji ofron VM-ve akses n\u00eb nj\u00eb pjes\u00eb t\u00eb kapaciteteve harduerike t\u00eb sistemit. Zgjidhja lejon ndarjen e procesorit grafik mes 16 p\u00ebrdoruesve me performanc\u00eb t\u00eb barabart\u00eb t\u00eb sistemeve t\u00eb virtualizuara.<\/p>\n

Sa i p\u00ebrket gjigantit t\u00eb dyt\u00eb IT, teknologjia e tyre bazohet<\/a><\/noindex> n\u00eb hypervisorin Citrix XenServer 7. Ajo bashkon funksionimin e drejtuesve standard t\u00eb GPU-ve dhe makinave virtuale, duke lejuar q\u00eb k\u00ebto t\u00eb fundit t\u00eb shfaqin aplikacione 3D dhe desktop\u00eb n\u00eb pajisjet e qindra p\u00ebrdoruesve.<\/p>\n

E ardhmja e teknologjis\u00eb<\/h2>\n

\nZhvilluesit e procesor\u00ebve grafik\u00eb virtual\u00eb b\u00ebjn\u00eb baste<\/a><\/noindex> n\u00eb integrimin e sistemeve AI dhe rritjen e popullaritetit t\u00eb zgjidhjeve me performanc\u00eb t\u00eb lart\u00eb n\u00eb tregun e teknologjive p\u00ebr biznes. Ata shpresojn\u00eb se nevoja p\u00ebr p\u00ebrpunimin e volumit t\u00eb madh t\u00eb t\u00eb dh\u00ebnave do t\u00eb rris\u00eb k\u00ebrkes\u00ebn p\u00ebr vGPU.<\/p>\n

Aktualisht, prodhuesit po k\u00ebrkojn\u00eb nj\u00eb m\u00ebnyr\u00eb<\/a><\/noindex> t\u00eb kombinohet funksionaliteti i CPU dhe GPU n\u00eb nj\u00eb b\u00ebrtham\u00eb, p\u00ebr t\u00eb p\u00ebrshpejtuar zgjidhjen e detyrave t\u00eb lidhura me grafik\u00ebn, kryerjen e llogaritjeve matematikore, operacioneve logjike dhe p\u00ebrpunimin e t\u00eb dh\u00ebnave. Shfaqja n\u00eb treg e k\u00ebtyre b\u00ebrthamave n\u00eb t\u00eb ardhmen do ta ndryshoj\u00eb qasjen ndaj virtualizimit t\u00eb burimeve dhe shp\u00ebrndarjes s\u00eb tyre mes ngarkesave punuese n\u00eb mjedisin virtual dhe n\u00eb re.<\/p>\n

\n\u00c7far\u00eb t\u00eb lexoni p\u00ebr tem\u00ebn n\u00eb blogun ton\u00eb t\u00eb korporat\u00ebs:<\/p>\n