Ana shirin yin amfani da tsarin Shenango a cibiyoyin bayanai.
/ hoto
A cewar daya daga cikin masu samar da bayanai, cibiyoyin bayanai kawai 20-40% na samuwa ikon kwamfuta. A babban lodi wannan nuna alama . Wannan rarraba albarkatun yana haifar da fitowar abin da ake kira "sabar sabar zombie". Waɗannan injuna ne waɗanda ke zama marasa aiki a mafi yawan lokuta, suna lalata makamashi. A yau 30% na sabobin a duniya , yana cinye dala biliyan 30 na wutar lantarki a kowace shekara.
MIT ta yanke shawarar yaƙar rashin ingantaccen amfani da albarkatun kwamfuta.
Ƙungiyar injiniya na'ura mai sarrafa kayan aiki mai suna Shenango. Manufarsa ita ce saka idanu akan yanayin ma'ajin aiki da sake rarraba hanyoyin da suka makale (wadanda ba za su iya karɓar lokacin CPU ba) zuwa injuna kyauta.
Yadda Shenango ke aiki
Shenango ɗakin karatu ne na Linux a cikin C tare da Rust da C++. Ana buga lambar aikin da aikace-aikacen gwaji a ciki ku GitHub.
Maganinta ya dogara ne akan algorithm IOKernel, wanda ke gudana akan ginshiƙi mai mahimmanci na tsarin multiprocessor. Yana sarrafa buƙatun CPU ta amfani da tsarin aiki , wanda ke ba da damar aikace-aikace don sadarwa kai tsaye tare da na'urorin cibiyar sadarwa.
IOKernel yana yanke shawarar waɗanne kernels don ba da takamaiman aiki zuwa. Algorithm kuma yana yanke shawarar adadin muryoyin da ake buƙata. Ga kowane tsari, an ƙaddara manyan abubuwan tsakiya (lamuni) da ƙarin (fashewa) - ana ƙaddamar da ƙarshen a yayin da ake haɓaka adadin buƙatun zuwa CPU.
An tsara jerin gwanon neman IOKernel azaman . Kowane dakika biyar biyar, algorithm ɗin yana bincika don ganin ko duk ayyukan da aka sanya wa ainihin sun kammala. Don yin wannan, yana kwatanta wurin halin yanzu na kan buffer da matsayi na baya na wutsiya. Idan ya bayyana cewa wutsiya ta riga ta kasance a cikin jerin gwano a lokacin rajistan da ya gabata, tsarin yana lura da nauyin buffer kuma yana ba da ƙarin mahimmanci don aiwatarwa.
Lokacin rarraba kaya, ana ba da fifiko ga maɓalli waɗanda aka aiwatar da tsarin iri ɗaya a kansu a baya kuma an rage wani yanki a cikin ma'ajin, ko ga kowane nau'in murhu marasa aiki.
Shenango kuma yana ɗaukar hanya . Matsakaicin da aka ware don gudanar da aikace-aikacen guda ɗaya suna lura da adadin ayyukan da juna ke da su. Idan daya cibiya ta gama jerin ayyukanta a gaban sauran, to tana “taimaka” wani ɓangare na kaya daga makwabta.
Abũbuwan amfãni da rashin amfani
By injiniyoyi daga MIT, Shenango yana da ikon sarrafa buƙatun miliyan biyar a sakan daya da kuma kiyaye matsakaicin lokacin amsawa na 37 microseconds. Masana sun ce a wasu lokuta fasahar na iya kara yawan amfani da na'urorin sarrafa bayanai a cibiyoyin bayanai zuwa kashi 100%. Sakamakon haka, ma'aikatan cibiyar bayanai za su iya yin ajiya akan siye da kula da sabar.
Yiwuwar Magani da kwararru daga wasu jami'o'i. A cewar wani farfesa daga wata cibiyar Koriya, tsarin MIT zai taimaka wajen rage jinkirin ayyukan yanar gizon. Misali, zai zama da amfani a shagunan kan layi. A kwanakin tallace-tallace akwai ma jinkiri na biyu a loda shafin don rage yawan ra'ayoyin rukunin yanar gizon da kashi 11%. Rarraba kaya da sauri zai taimaka wa ƙarin abokan ciniki hidima.
Har yanzu fasahar tana da kurakurai - ba ta goyan bayan multiprocessor -tsarin da aka haɗa kwakwalwan kwamfuta zuwa nau'ikan ƙwaƙwalwar ajiya daban-daban kuma ba sa “saɓani” da juna. A wannan yanayin, IOKernel na iya tsara aikin rukunin masu sarrafawa daban, amma ba duk kwakwalwan uwar garken ba.
/ hoto
Makamantan fasahohin
Sauran tsarin daidaita nauyin kayan sarrafawa sun haɗa da Arachne. Yana ƙididdige adadin muryoyi nawa aikace-aikacen zai buƙaci lokacin da ya fara, kuma yana rarraba matakai bisa ga wannan alamar. A cewar mawallafa, matsakaicin latency na aikace-aikacen a Arachne shine kusan 10 dubu micro seconds.
Ana aiwatar da fasahar azaman ɗakin karatu na C++ don Linux, kuma lambar tushe tana samuwa a .
Wani kayan aikin daidaitawa shine ZygOS. Kamar Shenango, fasahar tana amfani da hanyar satar aiki don sake rarraba matakai. A cewar mawallafin ZygOS, matsakaicin jinkirin aikace-aikacen lokacin amfani da kayan aiki shine kusan 150 micro seconds, kuma matsakaicin shine kusan 450 microsecond. Lambar aikin kuma .
binciken
Cibiyoyin bayanai na zamani na ci gaba da fadadawa, ana samun karuwar yanayin musamman a kasuwannin cibiyoyin bayanan hyperscale: yanzu a duniya. 430 hyperscale data cibiyoyin, amma a cikin shekaru masu zuwa adadin su na iya karuwa da 30%. A saboda wannan dalili, fasahar daidaita nauyin kayan sarrafawa za ta kasance cikin buƙatu mai girma. An riga an sami tsarin kamar Shenango manyan kamfanoni, kuma adadin irin waɗannan kayan aikin zai girma ne kawai a nan gaba.
Bugawa daga Kasuwancin Farko IaaS Blog:
source: www.habr.com