kupezeka kwa zinthu zokwanira zosungidwa pa ngozi;
pafupifupi makina deta ili pa centralized kapena kugawa dongosolo yosungirako;
kulekanitsa nthawi ya kayendetsedwe ka kayendetsedwe ka kayendetsedwe ka kayendetsedwe ka kayendetsedwe ka kayendetsedwe ka kayendetsedwe ka kayendetsedwe kake;
kusanja mkati mwa unyinji wa makasitomala omwe ali nawo;
kugwirizanitsa pokhapokha ngati pali kusalinganika kwakukulu, kusuntha kwa VM kothandiza kwambiri komanso kotetezeka (pambuyo pake, kusamuka kungalephereke);
kusanja makina amtundu wa "chete" (kusamuka kwa makina "aphokoso" kungatenge nthawi yayitali);
kulinganiza poganizira "mtengo" - katundu wosungirako ndi maukonde (ndi zomanga makonda kwa makasitomala akuluakulu);
kulinganiza poganizira za chikhalidwe cha munthu aliyense wa VM;
khalidwe la ogwiritsira ntchito machitidwe a chidziwitso cha kasitomala;
ma aligorivimu ogwiritsira ntchito ma seva a chidziwitso;
khalidwe la ma seva a DBMS;
katundu pazinthu zamakompyuta, makina osungira, maukonde;
kuyanjana kwa ma seva wina ndi mnzake polimbana ndi zida zamtambo.
Kuchulukitsitsa kwa ma seva ambiri ogwiritsira ntchito ndi ma database pazinthu zamtambo kumachitika pakapita nthawi, zotsatira zake zimatha kudziwonetsera ndikuphatikizana ndi zotsatira zosayembekezereka pakanthawi kosadziwika. Ngakhale kuwongolera njira zosavuta (mwachitsanzo, kuwongolera injini, makina otenthetsera madzi kunyumba), makina owongolera okha ayenera kugwiritsa ntchito zovuta. proportional-integral-kusiyana ma aligorivimu ndi mayankho.
Ntchito yathu ndi maulamuliro ambiri ovuta kwambiri, ndipo pali chiopsezo kuti dongosololi silingathe kulinganiza katunduyo kuzinthu zomwe zakhazikitsidwa mu nthawi yoyenera, ngakhale palibe zisonkhezero zakunja zochokera kwa ogwiritsa ntchito.
Mbiri ya zochitika zathu
Kuti tithetse vutoli, tinaganiza kuti tisayambe kuyambira pachiyambi, koma kulimbikitsa zomwe zilipo kale, ndipo tinayamba kuyanjana ndi akatswiri omwe ali ndi chidziwitso pa ntchitoyi. Mwamwayi, kumvetsetsa kwathu za vutoli kumagwirizana kwathunthu.
Gawo 1
Tidagwiritsa ntchito makina ozikidwa paukadaulo wa neural network ndikuyesa kukhathamiritsa zida zathu potengera izi.
Chidwi cha siteji iyi chinali kuyesa ukadaulo watsopano, ndipo kufunikira kwake kunali kugwiritsa ntchito njira yosagwirizana ndi njira yothetsera vuto pomwe, zinthu zina kukhala zofanana, njira zokhazikika zinali zitatopa.
Tinayambitsa dongosolo, ndipo tinayambadi kulinganiza. Kukula kwa mtambo wathu sikunatilole kuti tipeze zotsatira zabwino zonenedwa ndi omanga, koma zinali zoonekeratu kuti kusanja kumagwira ntchito.
Panthawi imodzimodziyo, tinali ndi zofooka zazikulu:
Kuti aphunzitse neural network, makina enieni amafunika kuyenda popanda kusintha kwakukulu kwa milungu kapena miyezi.
Algorithm idapangidwa kuti ikwaniritsidwe potengera kusanthula kwa "mbiri" zakale.
Kuphunzitsa neural network kumafuna kuchuluka kwa data ndi zida zamakompyuta.
Kukhathamiritsa ndi kusanja kungathe kuchitika kawirikawiri - kamodzi pa maola angapo, zomwe mwachidziwikire sizokwanira.
Choyamba - kwa makasitomala amakampani. Izi zikutanthauza kuti timafunikira dongosolo lomwe limagwira ntchito mwachangu, ndi zoletsa zamabizinesi zomwe zimangosavuta kukhazikitsa.
Funso lachiwiri - mukutanthauza chiyani ndi mawu oti "mwamsanga"? Chifukwa cha mkangano waufupi, tinaganiza kuti titha kuyamba ndi nthawi yoyankha kwa mphindi 5-10, kotero kuti maulendo afupikitsa ang'onoang'ono sangalowetse dongosololi mu resonance.
Funso lachitatu - ndi kukula kwanji kwa ma seva oyenera kusankha?
Nkhaniyi inathetsedwa yokha. Nthawi zambiri, makasitomala samapanga masanjidwe a seva kukhala akulu kwambiri, ndipo izi zikugwirizana ndi malingaliro a nkhaniyi kuti achepetse kuphatikizika kwa ma seva 30-40.
Kuphatikiza apo, pogawa magawo a seva, timachepetsa ntchito ya algorithm yolinganiza.
Funso lachinayi - Kodi maukonde a neural ndi oyenera bwanji kwa ife ndi njira yayitali yophunzirira komanso kusanja kosowa? Tinaganiza zosiya kuti tigwiritse ntchito ma aligorivimu osavuta kuti tipeze zotsatira mumasekondi.
Kufotokozera kwa dongosolo lomwe limagwiritsa ntchito ma aligorivimu oterowo ndi zovuta zake zitha kupezeka apa
Zikuwoneka kuti wina atha kukhazika mtima pansi pa izi, kudikirira kutsimikizika kotsimikizika ndikutseka mutuwo.
Koma timakakamizika kuchita gawo latsopano ndi mwayi wotsatirawu wodziwikiratu
Ziwerengero mwachitsanzo, apa ΠΈ apa zikuwonetsa kuti makina a purosesa awiri ndi anayi ndi otsika kwambiri pakuchitapo kanthu kuposa makina a purosesa imodzi. Izi zikutanthauza kuti ogwiritsa ntchito onse amalandira zotulutsa zochepa kwambiri kuchokera ku CPU, RAM, SSD, LAN, FC zogulidwa m'makina opangira ma multiprocessor poyerekeza ndi machitidwe amodzi.
Okonza zothandizira pawokha akhoza kukhala ndi zolakwika zazikulu, nayi imodzi mwankhani pa mutu uwu.
Tekinoloje zoperekedwa ndi Intel ndi AMD zowunikira RAM ndi cache zimapangitsa kuti zitheke kuphunzira momwe makina amagwirira ntchito ndikuwayika m'njira yoti oyandikana nawo "aphokoso" asasokoneze makina "achete".
Kukula kwa magawo (maukonde, makina osungira, choyambirira cha makina enieni, mtengo wakusamuka, kukonzekera kwake kusamuka).
Chiwerengero
Chotsatira cha ntchito yathu kuwongolera ma aligorivimu anali kunena momveka bwino kuti kugwiritsa ntchito ma aligorivimu amakono ndizotheka kukwaniritsa kukhathamiritsa kwakukulu kwazinthu zapakati pa data (25-30%) komanso nthawi yomweyo kupititsa patsogolo ntchito zamakasitomala.
Ma algorithm ozikidwa pa neural network ndi yankho losangalatsa, koma lomwe likufunika kupititsa patsogolo, ndipo chifukwa cha malire omwe alipo, siloyenera kuthana ndi vuto lamtunduwu pama voliyumu omwe amafanana ndi mitambo yachinsinsi. Panthawi imodzimodziyo, algorithm inasonyeza zotsatira zabwino mu mitambo yapagulu ya kukula kwakukulu.
Tikuwuzani zambiri za kuthekera kwa mapurosesa, ma scheduler, ndi kusanja kwapamwamba m'nkhani zotsatirazi.