Rodha Kuyera muOpenstack (Chikamu 2)

В chinyorwa chekupedzisira takataura nezve kuedza kwedu kushandisa Watcher uye takapa mushumo webvunzo. Isu tinopota tichiitisa bvunzo dzakadai dzekuenzanisa uye mamwe mabasa akakosha ebhizinesi hombe kana gore revashandisi.

Kuoma kwepamusoro kwedambudziko riri kugadziriswa zvingada zvinyorwa zvakawanda kutsanangura chirongwa chedu. Nhasi tiri kuburitsa chinyorwa chechipiri munhevedzano, yakatsaurirwa kuenzanisa machina chaiwo ari mugore.

Mamwe mazwi

Iyo kambani yeVmWare yakaunza iyo DRS (Distributed Resource Scheduler) utility kuenzanisa mutoro weiyo virtualization nharaunda yavakagadzira nekupa.

Sekunyora searchvmware.techtarget.com/definition/VMware-DRS
"VMware DRS (Distributed Resource Scheduler) chishandiso chinoyeresa computing mitoro nezviwanikwa zviripo munzvimbo chaiyo. Izvo zvinoshandiswa chikamu cheiyo virtualization suite inonzi VMware Infrastructure.

NeVMware DRS, vashandisi vanotsanangura mitemo yekugovera zviwanikwa zvemuviri pakati pemashini chaiwo (VMs). Iyo yekushandisa inogona kugadziridzwa kune manual kana otomatiki kudzora. VMware zviwanikwa madziva anogona kuwedzerwa nyore, kubviswa, kana kurongwazve. Kana zvichidikanwa, madziva ezvishandiso anogona kuparadzaniswa pakati pezvikamu zvakasiyana zvebhizinesi. Kana mutoro webasa pane mumwe kana anopfuura chaiwo muchina uchichinja zvakanyanya, VMware DRS inogoverazve mashini chaiwo pamasevha emuviri. Kana basa rese radzikira, mamwe maseva emuviri anogona kutorwa kunze kwenyika kwenguva pfupi uye basa rakabatanidzwa."

Nei kuenzanisa kuchidiwa?

Mune maonero edu, DRS inofanirwa-ine makore, kunyange izvi zvisingareve kuti DRS inofanira kushandiswa nguva dzose uye kwose kwose. Zvichienderana nechinangwa uye zvinodiwa zvegore, panogona kunge paine zvinodiwa zvakasiyana zveDRS uye nzira dzekuenzanisa. Panogona kuva nemamiriro ezvinhu apo kuenzana hakudiwi zvachose. Kana kutokuvadza.

Kuti unzwisise zviri nani kuti ndeipi uye ndeipi vatengi DRS inodiwa, ngatitarisei zvinangwa nezvinangwa zvavo. Makore anogona kukamurwa kuita zveruzhinji uye zvakavanzika. Heano misiyano mikuru pakati pemakore aya uye zvinangwa zvevatengi.

Yakavanzika makore / Makuru bhizinesi vatengi
Veruzhinji makore / Epakati uye mabhizinesi madiki, vanhu

Chiyero chikuru uye zvinangwa zvemushandisi
Kupa sevhisi yakavimbika kana chigadzirwa
Kuderedza mutengo webasa mukurwa mumusika wemakwikwi

Zvido zvebasa
Kuvimbika pamatanho ese uye mune ese system zvinhu

Yakavimbiswa kuita

Isa pamberi machina chaiwo mumapoka akati wandei 

Ruzivo uye kuchengetedzwa kwedata data

SLA uye XNUMX/XNUMX rutsigiro
Kunyanya kureruka kwekugamuchira sevhisi

Zvikati zviri nyore masevhisi

Basa re data rine mutengi

Hapana kukoshesa kweVM kunodiwa

Kuchengetedzwa kweruzivo pamwero weakajairika masevhisi, mutoro pamutengi

Panogona kunge paine glitches

Kwete SLA, mhando haina kuvimbiswa

Email rutsigiro

Backup haidiwi

Client Features
Yakakura kwazvo siyana yemaapplication.

Nhaka zvikumbiro zvakagara nhaka mukambani.

Yakaoma tsika zvivakwa zvemutengi wega wega.

Affinity mitemo.

Iyo software inoshanda isina kumira mu7x24 modhi. 

On-the-fly backup zvishandiso.

Zvinofungidzirwa cyclic mutengi mutoro.
Yakajairika maapplication - network kuenzanisa, Apache, WEB, VPN, SQL

Chikumbiro chinogona kumira kwechinguva

Inobvumira kugoverwa kusiri pamutemo kweVM mumakore

Mutengi backup

Inofanotaurwa statistically avhareji mutoro nenhamba huru yevatengi.

Zvinoreva zvekuvaka
Geoclustering

Centralized kana kugoverwa kuchengetedza

Reserved IBS
Nzvimbo yekuchengetedza data pamakomputa node

Kuenzanisa Zvinangwa
Kunyange kugovera mutoro

Maximum application inopindura 

Kashoma kunonoka nguva yekuenzanisa

Kuenzanisa chete pazvinenge zvakakodzera

Kuburitsa zvimwe zvekushandisa kuitira kudzivirira kuchengetedza
Kuderedza mari yebasa uye mutengo wevashandisi 

Kudzima zvimwe zviwanikwa kana zviine mutoro wakaderera

Kuchengetedza simba

Kuderedza mari yevashandi

Isu tinotora mhedzisiro inotevera isu pachedu:

Zvemakore egayakapihwa kune vatengi vemakambani makuru, DRS inogona kushandiswa pasi pezvirambidzo zvinotevera:

• kuchengetedzwa kweruzivo uye kufunga nezve affinity mitemo pakuenzanisa;
• kuvapo kwezviwanikwa zvakakwana zvakachengetwa pakaitika tsaona;
• virtual muchina data iri panzvimbo yepakati kana yakagoverwa yekuchengetedza system;
• nguva yekuparadzaniswa kwekutonga, kuchengetedza uye kuenzanisa maitiro;
• kuenzanisa chete mukati meaggregate yevatengi vanogamuchira;
• kuenzanisa chete kana paine kusaenzana kwakasimba, iyo inoshanda uye yakachengeteka VM kutama (mushure mezvose, kutama kunogona kukundikana);
• kuenzanisa michina yemhando "yakanyarara" (kutama kwemachina "ane ruzha" anogona kutora nguva yakareba kwazvo);
• kuenzanisa uchifunga nezve "mutengo" - mutoro pane yekuchengetedza sisitimu uye network (ine yakagadziridzwa zvivakwa zvevatengi vakakura);
• kuenzanisa uchifunga nezvemaitiro ega ega eVM;
• Kuenzanisa kunonyanya kuitwa panguva dzisiri dzekushanda (husiku, kupera kwevhiki, mazororo).

Zvemakore eruzhinjikupa masevhisi kune vatengi vadiki, DRS inogona kushandiswa kazhinji kazhinji, nehunyanzvi hwepamberi:

• kusavapo kwezvirambidzo zvekuchengetedza ruzivo uye mitemo yekubatana;
• kuyera mukati megore;
• kuenzanisa panguva ipi zvayo yakakodzera;
• kuenzanisa chero VM;
• kuenzanisa "ruzha" chaiwo muchina (kuitira kuti usakanganise vamwe);
• virtual machine data inowanzowanikwa pama disks emunharaunda;
• uchifunga nezveavhareji yekushanda kwekuchengetedza masisitimu uye network (iyo cloud architecture yakabatana);
• kuenzanisa maererano nemitemo yakajairika uye inowanikwa data center maitiro maitiro.

Kuoma kwedambudziko

Kuoma kwekuenzanisa ndeyekuti DRS inofanira kushanda nehuwandu hwezvinhu zvisina chokwadi:

• maitiro evashandisi vega yega yevatengi 'ruzivo masisitimu;
• algorithms yekushanda kwemaseva ehurongwa hwemashoko;
• maitiro emaseva eDBMS;
• mutoro pamakomputa zviwanikwa, masisitimu ekuchengetedza, network;
• kudyidzana kwemaseva kune mumwe nemumwe mukurwira kwegore zviwanikwa.

Kuremerwa kwenhamba yakawanda yemaseva ekushandisa uye dhatabhesi pazviwanikwa zvegore zvinoitika nekufamba kwenguva, mhedzisiro inogona kuzviratidza uye kudhumhana nemhedzisiro isingatarisike pane imwe nguva isingafungidzirwe. Kunyangwe kudzora maitiro akareruka (semuenzaniso, kudzora injini, mvura yekudziya sisitimu kumba), otomatiki kudzora masisitimu anoda kushandisa yakaoma. proportional-integral-differentiating algorithms nemhinduro.

Basa redu mirairo yakawanda yehukuru yakanyanya kuomarara, uye pane njodzi yekuti sisitimu haizokwanisa kuenzanisa mutoro kune dzakasimbiswa hunhu munguva inonzwisisika, kunyangwe pasina ekunze pesvedzero kubva kuvashandisi.

Nhoroondo yezvakaitika kwatiri

Kuti tigadzirise dambudziko iri, takasarudza kusatanga kubva pakutanga, asi kuvaka pane zvakaitika kare, uye takatanga kusangana nenyanzvi dzine ruzivo mundima iyi. Sezvineiwo, kunzwisisa kwedu kwedambudziko kwakanyatsoenderana.

Stage 1

Isu takashandisa system yakavakirwa neural network tekinoroji uye takaedza kukwenenzvera zviwanikwa zvedu zvichibva pairi.

Kufarirwa kweichi nhanho kwaive mukuyedza tekinoroji nyowani, uye kukosha kwayo kwaive mukushandisa nzira isiri-yakajairwa kugadzirisa dambudziko apo, zvimwe zvinhu zvakaenzana, nzira dzakajairwa dzakange dzazvipedza simba.

Takatanga hurongwa, uye takatanga kunyatsoita kuenzanisa. Huyero hwegore redu hahuna kutitendera kuti tiwane tarisiro yakataurwa nevagadziri, asi zvaive pachena kuti kuenzanisa kwaive kushanda.

Panguva imwe cheteyo, takanga tine ganhuriro dzakakomba zvikuru:

• Kudzidzisa neural network, machina chaiwo anoda kumhanya pasina shanduko yakakosha kwemavhiki kana mwedzi.
• Iyo algorithm yakagadzirirwa optimization zvichienderana nekuongororwa kwekare "nhoroondo" data.
• Kudzidzisa neural network kunoda huwandu hwakati wandei hwe data uye zviwanikwa zvekombuta.
• Kugadzirisa uye kuenzanisa kunogona kuitwa kashoma - kamwe chete maawa mashoma, izvo zviri pachena kuti hazvina kukwana.

Stage 2

Sezvo isu tisina kugutsikana nemamiriro ezvinhu, takasarudza kugadzirisa hurongwa, uye kuita izvi, mhinduro mubvunzo mukuru – tiri kugadzirira ani?

Kutanga - kune vatengi vemakambani. Izvi zvinoreva kuti tinoda system inoshanda nekukurumidza, ine zvirambidzo zvemakambani zvinongorerutsa kuita.

Mubvunzo wechipiri - unorevei nezwi rokuti "nekukurumidza"? Somugumisiro wekukakavadzana kwenguva pfupi, takasarudza kuti tinogona kutanga nenguva yekupindura ye5-10 maminetsi, kuitira kuti kuvhiyiwa kwenguva pfupi kurege kuunza hurongwa mu resonance.

Mubvunzo wechitatu - ndeipi saizi yenhamba yakaenzana yemaseva ekusarudza?
Nyaya iyi yakazvigadzirisa. Kazhinji, vatengi havaite maseva aggregations yakakura kwazvo, uye izvi zvinoenderana nekurudziro yechinyorwa kudzikamisa kuunganidzwa kune makumi matatu kusvika makumi mana maseva.

Uye zvakare, nekupatsanura server pool, isu tinorerutsa basa reiyo balancing algorithm.

Mubvunzo wechina - Yakakodzera sei neural network kwatiri nemaitiro ayo ekudzidza kwenguva refu uye isingawanzo kuenzanisa? Isu takasarudza kuisiya tichifarira maalgorithms akareruka kuitira kuti tiwane mhinduro mumasekonzi.

Tsanangudzo yehurongwa hunoshandisa maalgorithms akadaro uye nekuipa kwayo kunogona kuwanikwa pano

Isu takaita uye nekutangisa iyi sisitimu uye takagamuchira zvinokurudzira - ikozvino inogara ichiongorora iyo gore kuremerwa uye inoita kurudziro yekufambisa chaiyo michina, iyo yakanyanya kunaka. Kunyangwe iye zvino zviri pachena kuti tinogona kuwana 10-15% kusunungurwa kwezviwanikwa zvemichina mitsva yemagetsi apo tichivandudza hutano hwebasa rezviripo.

Kana kusaenzana mu RAM kana CPU kwaonekwa, iyo sisitimu inoraira kune Tionix scheduler kuti aite kutama kwehupenyu hwemakina anodiwa anodiwa. Sezvinoonekwa kubva kuhurongwa hwekutarisa, iyo chaiyo muchina wakafamba kubva kune imwe (yepamusoro) ichienda kune imwe (yepasi) mugadziri uye akasunungura ndangariro pane yepamusoro host (yakasimudzwa muyero madenderedzwa), zvichiteerana ichiitora pane yezasi (yakasimudzwa muchena. madenderedzwa).

Iye zvino tiri kuedza kunyatsoongorora kushanda kwegorgorithm yazvino uye tiri kuedza kutsvaga zvikanganiso zvinogona kuitika mairi.

Stage 3

Zvinoita sekuti munhu anogona kudzikama pane izvi, kumirira kushanda kwakaratidza uye kuvhara musoro.
Asi isu tinosundirwa kuita nhanho nyowani neinotevera iri pachena optimization mikana

1. Statistics, semuenzaniso, pano и pano inoratidza kuti maviri-uye mana-processor masisitimu akadzikira zvakanyanya mukuita pane imwechete-processor masisitimu. Izvi zvinoreva kuti vese vashandisi vanogamuchira zvakanyanya kushoma kuburitsa kubva kuCPU, RAM, SSD, LAN, FC yakatengwa mumamultiprocessor masisitimu achienzaniswa neamwechete-processor masisitimu.
2. Vagadziri vezvishandiso pachavo vanogona kunge vaine zvikanganiso zvakakomba, hechino chimwe chezvinyorwa pamusoro penyaya iyi.
3. Tekinoroji dzinopihwa neIntel ne AMD yekutarisa RAM uye cache inoita kuti zvikwanise kudzidza maitiro emakina chaiwo uye nekuaisa nenzira yekuti "ine ruzha" vavakidzani havakanganise "yakanyarara" chaiwo muchina.
4. Kuwedzera kweseti yemaparamita (network, kuchengetedza system, kukoshesa kweiyo chaiyo muchina, mutengo wekufamba, kugadzirira kwayo kutama).

Total

Mhedzisiro yebasa redu rekuvandudza kuenzanisa maalgorithms yaive mhedziso yakajeka yekuti kushandisa algorithms yemazuva ano zvinogoneka kuwana yakakosha optimization ye data center zviwanikwa (25-30%) uye panguva imwe chete kuvandudza kunaka kwebasa revatengi.

Algorithm yakavakirwa pane neural network zvirokwazvo mhinduro inonakidza, asi inoda imwe kuvandudzwa, uye nekuda kwezvipimo zviripo, haina kukodzera kugadzirisa rudzi urwu rwedambudziko pamavhoriyamu akajairwa nemakore akavanzika. Panguva imwecheteyo, iyo algorithm yakaratidza mhedzisiro yakanaka mumakore eruzhinji ehukuru hwakakosha.

Isu tichakuudza zvimwe nezve kugona kwevagadziri, vanoronga, uye yakakwirira-level kuenzanisa mune zvinotevera zvinyorwa.

Source: www.habr.com