Π sikhulume ngemizamo yethu yokusebenzisa i-Watcher futhi sanikeza umbiko wokuhlola. Isikhathi ngasinye senza izivivinyo ezinjalo zokulinganisa neminye imisebenzi ebalulekile yebhizinisi elikhulu noma ifu lomsebenzisi.
Ubunkimbinkimbi obuphezulu benkinga exazululwayo bungase budinge izindatshana ezimbalwa ukuchaza iphrojekthi yethu. Namuhla sishicilela isihloko sesibili ochungechungeni, esinikezelwe ekulinganiseni imishini ebonakalayo emafini.
Amanye amatemu
Inkampani ye-VmWare yethule insiza ye-DRS (Distributed Resource Scheduler) ukuze ilinganisele umthwalo wendawo yokwenza izinto ezibonakalayo abayithuthukisile futhi bayinikeza.
Njengoba ebhala
βI-VMware DRS (Isihleli Sensiza Esisabalalisiwe) siwuhlelo olulinganisa imithwalo yekhompyutha nezinsiza ezitholakalayo endaweni ebonakalayo. Uhlelo lokusebenza luyingxenye ye-virtualization suite ebizwa nge-VMware Infrastructure.
Nge-VMware DRS, abasebenzisi bachaza imithetho yokusabalalisa izinsiza ezibonakalayo phakathi kwemishini ebonakalayo (ama-VM). Uhlelo lokusebenza lungalungiselelwa ukulawula okwenziwa ngesandla noma okuzenzakalelayo. Amachibi ezinsiza ze-VMware angangezwa kalula, asuswe, noma ahlelwe kabusha. Uma efisa, amachibi ezinsiza angahlukaniswa phakathi kwamayunithi ebhizinisi ahlukene. Uma umthwalo womsebenzi emshinini owodwa noma eminingi ebonakalayo ushintsha kakhulu, i-VMware DRS isabalalisa kabusha imishini ebonakalayo kuwo wonke amaseva aphathekayo. Uma wonke umsebenzi wehla, amanye amaseva aphathekayo angase akhishwe okwesikhashana angaxhunyiwe ku-inthanethi futhi nomsebenzi uhlanganiswe."
Kungani kudingeka ukulinganisa?
Ngokombono wethu, i-DRS iyisici okumelwe sibe nayo sefu, nakuba lokhu akusho ukuthi i-DRS kufanele isetshenziswe njalo futhi yonke indawo. Kuye ngenjongo nezidingo zefu, kungase kube nezidingo ezihlukile ze-DRS nezindlela zokulinganisa. Kungase kube nezimo lapho ukulinganisa kungadingeki nhlobo. Noma yingozi.
Ukuze uqonde kangcono ukuthi i-DRS idingeka kuphi futhi imaphi amaklayenti, ake sicabangele imigomo nezinhloso zawo. Amafu angahlukaniswa abe umphakathi kanye nangasese. Nawu umehluko omkhulu phakathi kwalawa mafu nezinjongo zamakhasimende.
Amafu angasese / amaklayenti ebhizinisi elikhulu
Amafu omphakathi / Amabhizinisi Aphakathi namancane, abantu
Umbandela oyinhloko kanye nemigomo yomsebenzisi
Ukunikeza isevisi ethembekile noma umkhiqizo
Ukunciphisa izindleko zezinsizakalo ekulweni emakethe yokuncintisana
Izidingo zesevisi
Ukuthembeka kuwo wonke amaleveli nakuzo zonke izici zesistimu
Ukusebenza okuqinisekisiwe
Beka kuqala imishini ebonakalayo ibe yizigaba ezimbalwa
Ulwazi nokuvikeleka kwedatha ebonakalayo
I-SLA kanye nokwesekwa okungu-XNUMX/XNUMX
Ubulula obukhulu bokuthola isevisi
Amasevisi alula uma kuqhathaniswa
Isibopho sedatha sikuye iklayenti
Akukho ukubeka phambili kwe-VM okudingekayo
Ukuphepha kolwazi ezingeni lezinsizakalo ezijwayelekile, umthwalo wemfanelo kuklayenti
Kungase kube namaphutha
Ayikho i-SLA, ikhwalithi ayiqinisekisiwe
Usekelo lwe-imeyili
Ukwenza ikhophi yasenqolobaneni akudingekile
Izici Zeklayenti
Uhla olubanzi kakhulu lwezinhlelo zokusebenza.
Izicelo zefa ezizuzwe njengefa enkampanini.
Izakhiwo eziyinkimbinkimbi zangokwezifiso zeklayenti ngalinye.
Imithetho yobudlelwano.
Isoftware isebenza ngaphandle kokuma kumodi engu-7x24.
Amathuluzi okulondoloza lapho undiza.
Umthwalo wekhasimende owumjikelezo obikezelwayo.
Izinhlelo zokusebenza ezijwayelekile - ukulinganisa kwenethiwekhi, i-Apache, i-WEB, i-VPN, i-SQL
Isicelo singama isikhashana
Ivumela ukusatshalaliswa ngokungafanele kwama-VM efwini
Ikhophi yasenqolobaneni yeklayenti
Umthwalo olinganiselwe ngokwezibalo onenani elikhulu lamaklayenti.
Imithelela yezokwakha
I-Geoclustering
Isitoreji esimaphakathi noma esabalalisiwe
I-IBS egodliwe
Ukugcinwa kwedatha yendawo kumanodi wokubala
Ukulinganisa Imigomo
Ngisho nokusatshalaliswa komthwalo
Ukusabela okuphezulu kohlelo lokusebenza
Isikhathi esincane sokulibaziseka sokulinganisa
Ukulinganisa kuphela uma kudingeka ngokucacile
Ukukhipha amathuluzi athile ukuze kugcinwe ukuzivikela
Ukunciphisa izindleko zesevisi kanye nezindleko zomqhubi
Ikhubaza ezinye izinsiza uma umthwalo uphansi
Ukonga Amandla
Ukunciphisa izindleko zabasebenzi
Sizenzela iziphetho ezilandelayo:
Okwamafu angaseseinikezwe amakhasimende amakhulu ezinkampani, i-DRS ingasetshenziswa ngaphansi kwemikhawulo elandelayo:
- ukuphepha kolwazi kanye nokucabangela imithetho yokuhambisana lapho kulinganisa;
- ukutholakala kwezinsiza ezanele ezibekiwe uma kwenzeka ingozi;
- idatha yomshini obonakalayo itholakala kusistimu yokugcina indawo eyodwa noma esabalalisiwe;
- ukuphatha okumangalisayo, izinqubo zokusekelayo nokulinganisa ngokuhamba kwesikhathi;
- ukulinganisa kuphela phakathi kweqoqo labasingathi bamakhasimende;
- ukulinganisa kuphela uma kukhona ukungalingani okuqinile, ukufuduka kwe-VM okuphumelelayo kakhulu nokuphephile (ngemuva kwakho konke, ukufuduka kungahluleka);
- ukulinganisa imishini ebonakalayo "ethule" (ukufuduka kwemishini ebonakalayo "enomsindo" kungathatha isikhathi eside kakhulu);
- ukulinganisa kucabangela "izindleko" - umthwalo ohlelweni lwesitoreji nenethiwekhi (enezakhiwo ezenziwe ngokwezifiso zamaklayenti amakhulu);
- ukulinganisa kucatshangelwa izici zokuziphatha komuntu ngamunye we-VM ngayinye;
- Ukulinganisa kwenziwa ngokungcono ngezikhathi ezingezona ezokusebenza (ubusuku, izimpelasonto, amaholide).
Okwamafu omphakathiukuhlinzeka ngezinsizakalo kumakhasimende amancane, i-DRS ingasetshenziswa kaningi, ngamakhono athuthukile:
- ukungabi khona kwemikhawulo yokuphepha kolwazi kanye nemithetho yokuhambisana;
- ukulinganisa phakathi kwefu;
- ukulinganisa nganoma yisiphi isikhathi esifanele;
- ukulinganisa noma iyiphi i-VM;
- ukulinganisa imishini ebonakalayo "enomsindo" (ukuze ingaphazamisi abanye);
- idatha yomshini obonakalayo ivame ukutholakala kumadiski endawo;
- kucatshangelwa ukusebenza okumaphakathi kwezinhlelo zesitoreji namanethiwekhi (isakhiwo samafu sinobunye);
- ukulinganisa ngokuya ngemithetho ejwayelekile kanye nezibalo zokuziphatha zesikhungo sedatha.
Ukuba yinkimbinkimbi kwenkinga
Ubunzima bokulinganisa ukuthi i-DRS kufanele isebenze nenani elikhulu lezici ezingaqinisekile:
- ukuziphatha kwabasebenzisi bohlelo lolwazi lwekhasimende ngalinye;
- ama-algorithms okusebenza kwamaseva esistimu yolwazi;
- ukuziphatha kwamaseva e-DBMS;
- layisha izinsiza zekhompyutha, izinhlelo zokugcina, inethiwekhi;
- ukusebenzisana kwamaseva komunye nomunye emzabalazweni wezinsiza zamafu.
Umthwalo wenombolo enkulu yamaseva wohlelo lokusebenza olubonakalayo kanye nemininingwane yolwazi ezinsizeni zamafu kwenzeka ngokuhamba kwesikhathi, imiphumela ingaziveza futhi idlulele komunye nomunye ngomphumela ongalindelekile ngesikhathi esingalindelekile. Ngisho nokulawula izinqubo ezilula (isibonelo, ukulawula injini, uhlelo lokushisisa amanzi ekhaya), amasistimu okulawula okuzenzakalelayo kudingeka asebenzise inkimbinkimbi. ama-algorithms anempendulo.
Umsebenzi wethu unemiyalo eminingi yobukhulu eyinkimbinkimbi kakhulu, futhi kunengozi yokuthi uhlelo ngeke lukwazi ukulinganisa umthwalo kumanani amisiwe ngesikhathi esifanele, noma ngabe awekho amathonya angaphandle avela kubasebenzisi.
Umlando wentuthuko yethu
Ukuze sixazulule le nkinga, sinqume ukuthi singaqali kusukela ekuqaleni, kodwa sakhe phezu kokuhlangenwe nakho okukhona, futhi saqala ukuxhumana nochwepheshe abanolwazi kule ndawo. Ngenhlanhla, ukuqonda kwethu le nkinga kwaqondana ngokuphelele.
Isigaba 1
Sisebenzise isistimu esekelwe kubuchwepheshe benethiwekhi ye-neural futhi sazama ukuthuthukisa izinsiza zethu ngokusekelwe kuyo.
Intshisekelo yalesi sigaba kwakuwukuhlola ubuchwepheshe obusha, futhi ukubaluleka kwayo kwakuwukusebenzisa indlela engajwayelekile ekuxazululeni inkinga lapho, ezinye izinto zilingana, izindlela ezijwayelekile zazizikhandle ngokwazo.
Sethule uhlelo, futhi ngempela saqala ukulinganisa. Isikali sefu lethu asizange sisivumele ukuthi sithole imiphumela enethemba eshiwo onjiniyela, kodwa bekusobala ukuthi ukulinganisa kwakusebenza.
Ngesikhathi esifanayo, sasinemikhawulo engathi sΓna:
- Ukuze uqeqeshe inethiwekhi ye-neural, imishini ebonakalayo idinga ukusebenza ngaphandle kwezinguquko ezibalulekile amaviki noma izinyanga.
- I-algorithm yakhelwe ukuthuthukiswa ngokusekelwe ekuhlaziyweni kwedatha "engokomlando" yangaphambili.
- Ukuqeqesha inethiwekhi ye-neural kudinga inani elikhulu kakhulu ledatha nezinsiza zekhompuyutha.
- Ukuthuthukisa nokulinganisa kungenziwa ngokungavamile - kanye njalo emahoreni ambalwa, okusobala ukuthi akwanele.
Isigaba 2
Njengoba sasingenelisekile ngesimo sezindaba, sanquma ukulungisa uhlelo, futhi ukwenza lokhu, siphendule umbuzo oyinhloko β sikwenzela bani?
Okokuqala - kumakhasimende ezinkampani. Lokhu kusho ukuthi sidinga uhlelo olusebenza ngokushesha, olunaleyo mikhawulo yezinkampani eyenza kuphela ukuqaliswa kube lula.
Umbuzo wesibili - usho ukuthini ngegama elithi "ngokushesha"? Njengomphumela wenkulumo-mpikiswano emfushane, sinqume ukuthi singaqala ngesikhathi sokuphendula semizuzu engu-5-10, ukuze ukuhlinzwa kwesikhashana kungeke kwethule uhlelo ku-resonance.
Umbuzo wesithathu β iyiphi isayizi yenani elilinganiselwe lamaseva okufanele ukhethe?
Lolu daba luzixazulule ngokwalo. Ngokuvamile, amaklayenti awakwenzi ukuhlanganiswa kweseva kube kukhulu kakhulu, futhi lokhu kuhambisana nezincomo ze-athikili zokukhawulela ukuhlanganisa kumaseva angu-30-40.
Ngaphezu kwalokho, ngokuhlukanisa i-pool pool, senza umsebenzi we-algorithm yokulinganisa ibe lula.
Umbuzo wesine - ifaneleka kangakanani inethiwekhi ye-neural kithi ngenqubo yayo ende yokufunda nokulinganisa okungajwayelekile? Sinqume ukuyiyeka ukuze sivumele ama-algorithms okusebenza alula ukuze sithole imiphumela ngemizuzwana.
Incazelo yesistimu esebenzisa ama-algorithms anjalo kanye nebubi bayo ingatholakala
Sisebenzise futhi sethula le sistimu futhi sathola imiphumela ekhuthazayo - manje isihlaziya njalo umthwalo wamafu futhi yenza izincomo zokuhambisa imishini ebonakalayo, elungile kakhulu. Ngisho namanje kusobala ukuthi singakwazi ukufeza ukukhululwa kwe-10-15% yezinsiza zemishini emisha ebonakalayo ngenkathi sithuthukisa ikhwalithi yomsebenzi waleyo ekhona.
Uma kutholwa ukungalingani ku-RAM noma i-CPU, isistimu ikhipha umyalo kumhleli we-Tionix ukuthi enze ukuthuthela bukhoma kwemishini ebonakalayo edingekayo. Njengoba kubonakala ohlelweni lokuqapha, umshini we-virtual wasuka komunye (ophezulu) uye komunye (ophansi) futhi wakhulula inkumbulo kumsingathi ophezulu (okuqokonyiswe emibuthanweni ephuzi), ngokulandelanayo ehlala kwephansi (egqanyiswe ngokumhlophe. imibuthano).
Manje sizama ukuhlola ngokunembe kakhudlwana ukusebenza kahle kwe-algorithm yamanje futhi sizama ukuthola amaphutha okungenzeka kuwo.
Isigaba 3
Kubonakala sengathi umuntu angakwazi ukuzola kulokhu, alinde ukusebenza okuqinisekisiwe futhi avale isihloko.
Kodwa siphoqeleka ukuthi senze isigaba esisha ngala mathuba alandelayo okuthuthukisa asobala
- Izibalo, isibonelo, ΠΈ ikhombisa ukuthi amasistimu amaphrosesa amabili namane aphansi kakhulu ekusebenzeni kunezinhlelo zeprosesa eyodwa. Lokhu kusho ukuthi bonke abasebenzisi bathola okukhiphayo okuncane kakhulu okuvela ku-CPU, i-RAM, i-SSD, i-LAN, i-FC ethengwe kumasistimu we-multiprocessor uma kuqhathaniswa nezinhlelo zokucubungula okukodwa.
- Abahleli bezinsiza ngokwabo bangase babe namaphutha amakhulu, kulesi sihloko.
- Ubuchwepheshe obuhlinzekwa yi-Intel ne-AMD bokuqapha i-RAM kanye nenqolobane kwenza kube nokwenzeka ukutadisha ukuziphatha kwemishini ebonakalayo futhi ibekwe ngendlela yokuthi omakhelwane "abanomsindo" bangaphazamisi imishini ebonakalayo "ethule".
- Ukunwetshwa kwesethi yemingcele (inethiwekhi, uhlelo lokugcina, okubalulekile komshini obonakalayo, izindleko zokufuduka, ukulungela kwayo ukufuduka).
Inani
Umphumela womsebenzi wethu wokuthuthukisa ama-algorithms okulinganisa kwaba isiphetho esicacile sokuthi ukusebenzisa ama-algorithms esimanje kungenzeka ukufeza ukuthuthukiswa okubalulekile kwezinsiza zesikhungo sedatha (25-30%) futhi ngesikhathi esifanayo sithuthukise ikhwalithi yenkonzo yamakhasimende.
I-algorithm esekelwe kumanethiwekhi e-neural ngokuqinisekile iyisixazululo esithakazelisayo, kodwa esidinga ukuthuthukiswa okwengeziwe, futhi ngenxa yemikhawulo ekhona, ayifaneleki ukuxazulula lolu hlobo lwenkinga kumavolumu ajwayelekile amafu angasese. Ngesikhathi esifanayo, i-algorithm ibonise imiphumela emihle emafwini omphakathi osayizi obalulekile.
Sizokutshela kabanzi ngamakhono abacubungula, abahleli beshejuli, nokulinganisa okusezingeni eliphezulu ezihlokweni ezilandelayo.
Source: www.habr.com