Musangano weHabr haisi nyaya yekutanga. Pakutanga, takaita zviitiko zvakakura zveToaster zve300-400 vanhu, asi zvino takasarudza kuti misangano midiki ine musoro ichave yakakosha, iyo nzira yaunogona kuisa, semuenzaniso, mumhinduro. Musangano wekutanga wefomati iyi wakaitwa muna Chikunguru uye wakatsaurirwa kusimudzira kumashure. Vatori vechikamu vakateerera mishumo yezvimiro zveshanduko kubva kumashure kuenda kuML uye nezve dhizaini yebasa reQuadrupel paState Services portal, uye zvakare vakatora chikamu mutafura yakatenderedzwa yakatsaurirwa kuServerless. Kune avo vasina kukwanisa kuenda kuchiitiko ichi pachavo, mune ino post tinokuudza zvinhu zvinonyanya kufadza.
Kubva kumashure kusimudzira kusvika pakudzidza muchina
Chii chinoita mainjiniya edata muML? Ndeapi mabasa emugadziri webackend uye ML injiniya akafanana uye akasiyana? Ndeipi nzira yaunofanirwa kutora kuti uchinje yako yekutanga basa kune yechipiri yako? Izvi zvakataurwa naAlexander Parinov, uyo akapinda muchina kudzidza mushure memakore gumi ebasa rekudzokera shure.
Alexander Parinov
Nhasi Alexander anoshanda semugadziri wemacomputer ekuona masisitimu paX5 Retail Group uye anobatsira kuOpen Source mapurojekiti ane chekuita nekuona komputa uye kudzidza kwakadzama (github.com/creafz). Unyanzvi hwake hunosimbiswa nekupinda kwake muchikamu chepamusoro chezana chepasi rose cheKaggle Master (kaggle.com/creafz), chikuva chakakurumbira chemakwikwi ekudzidza muchina.
Sei uchichinjira kune muchina kudzidza
Gore nehafu rapfuura, Jeff Dean, mukuru weGoogle Brain, Google yakadzika yekudzidza-yakavakirwa artificial artificial intelligence research project, yakatsanangura kuti hafu yemiriyoni mitsara yekodhi muGoogle Translate yakatsiviwa sei neTensor Flow neural network ine mitsara mazana mashanu chete. Mushure mekudzidzisa network, mhando yedata yakawedzera uye zvivakwa zvakava nyore. Zvingaita sekuti iyi ndiyo ramangwana redu rakajeka: hatichafaniri kunyora kodhi, zvakakwana kuita neurons uye kuazadza nedata. Asi mukuita zvinhu zvose zvakanyanya kuoma.
ML zvivakwa paGoogle
Neural network ingori chikamu chidiki chezvivakwa (iyo diki dema mativi pamufananidzo uri pamusoro). Mazhinji mamwe masisitimu ebetsero anodiwa kuti ugamuchire data, kuigadzira, kuichengeta, kutarisa mhando, nezvimwe, tinoda zvivakwa zvekudzidziswa, kuendesa kodhi yemuchina yekudzidza mukugadzira, uye kuyedza iyi kodhi. Ese aya mabasa akafanana chaizvo neanoitwa backend Developers.
Machine kudzidza maitiro
Ndeupi musiyano uripo pakati peML uye backend?
Mune classical programming, tinonyora kodhi uye izvi zvinoraira maitiro echirongwa. MuML, isu tine diki modhi kodhi uye yakawanda data yatinokanda pamuenzaniso. Dhata muML yakakosha zvikuru: iyo imwechete modhi yakadzidziswa pane dzakasiyana data inogona kuratidza mhedzisiro dzakasiyana. Dambudziko nderekuti iyo data inogara yakapararira uye kuchengetwa mune akasiyana masisitimu (relational dhatabhesi, NoSQL dhatabhesi, matanda, mafaera).
Data shanduro
ML inoda kushandura kwete chete kodhi, semukukura kwechinyakare, asiwo iyo data: zvinodikanwa kuti unzwisise zvakajeka kuti muenzaniso wakadzidziswa chii. Kuita izvi, unogona kushandisa yakakurumbira Data Science Version Control raibhurari (dvc.org).
Data markup
Basa rinotevera nderekunyora data. Semuyenzaniso, maka zvinhu zvese zviri pamufananidzo kana kutaura kuti ndewekirasi ipi. Izvi zvinoitwa nemasevhisi akakosha seYandex.Toloka, iro basa rinorerutswa zvakanyanya nekuvapo kweAPI. Matambudziko anomuka nekuda kwe "munhu chinhu": unogona kuvandudza mhando yedata uye kuderedza zvikanganiso kusvika padiki nekupa basa rimwechete kune vakati wandei vanoita.
Kuonekwa muTensor Board
Kutema zviedzo kwakakosha kuenzanisa mibairo uye sarudza yakanakisa modhi zvichibva pane mamwe metrics. Kune yakakura seti yezvishandiso zvekuona - semuenzaniso, Tensor Board. Asi hapana nzira dzakanakisa dzekuchengeta zviedzo. Makambani madiki anowanzo kuita neExcel spreadsheet, nepo makuru anoshandisa akakosha mapuratifomu ekuchengetedza mhedzisiro mudhatabhesi.
Kune akawanda mapuratifomu ekudzidza muchina, asi hapana kana imwe yadzo inovhara 70% yezvinodiwa
Dambudziko rekutanga iro munhu anofanirwa kutarisana naro kana achiisa akadzidziswa modhi mukugadzira rine chekuita neanofarira chishandiso che data masayendisiti - Jupyter Notebook. Iko hakuna modularity mairi, ndiko kuti, iyo inobuda ndeye "tsoka" yekodhi yekodhi iyo isina kukamurwa kuita zvidimbu zvine musoro - modules. Zvose zvakasanganiswa: makirasi, mabasa, zvigadziro, nezvimwewo. Iyi code yakaoma kushandura uye kuedza.
Ndoita sei neizvi? Iwe unogona kuzvisiya, seNetflix, uye gadzira yako chikuva chinokutendera kuti utange aya malaptop zvakananga mukugadzira, endesa data kwavari sekuisa uye kuwana mhinduro. Iwe unogona kumanikidza vanogadzira vari kutenderedza modhi mukugadzira kuti vanyorezve kodhi kazhinji, vachiityora kuita mamodule. Asi nenzira iyi zviri nyore kuita chikanganiso, uye modhi yacho haishande sezvaidiwa. Naizvozvo, iyo yakanakira sarudzo ndeyekurambidza kushandiswa kweJupyter Notebook yemhando kodhi. Kana, chokwadi, data masayendisiti anobvumirana neizvi.
Muenzaniso sebhokisi dema
Nzira iri nyore yekuwana modhi mukugadzirwa ndeyekuishandisa sebhokisi dema. Iwe une imwe mhando yemhando yekirasi yemuenzaniso, iwe wakapihwa huremu hwemuenzaniso (maparamendi emaneuroni etiweki akadzidziswa), uye kana ukatanga kirasi iyi (fonera iyo nzira yekufungidzira, idyise mufananidzo), uchawana imwe. kufanotaura sechibuda. Zvinoitika mukati hazvina basa.
Kuparadzanisa server maitiro nemuenzaniso
Iwe unogona zvakare kusimudza imwe nzira yakaparadzana woitumira kuburikidza neRPC queue (ine mifananidzo kana imwe nzvimbo yedata. Pazvinobuda tinogashira kufanotaura.
Muenzaniso wekushandisa modhi muFlask:
@app.route("/predict", methods=["POST"])
def predict():
image = flask.request.files["image"].read()
image = preprocess_image(image)
predictions = model.predict(image)
return jsonify_prediction(predictions)
Dambudziko nemaitiro aya nderekuganhurwa kwekuita. Ngatitii isu tine Phyton kodhi yakanyorwa nemasainzi data inononoka, uye isu tinoda kudzvanya kunze kwekuita kwakanyanya. Kuti uite izvi, unogona kushandisa zvishandiso zvinoshandura iyo kodhi kuita yekuzvarwa kana kuishandura kuita imwe dhizaini yakagadzirirwa kugadzirwa. Kune maturusi akadaro kune ese magadzirirwo, asi hapana akanakisa; iwe unofanirwa kuwedzera iwo iwe pachako.
Izvo zvivakwa muML zvakafanana neyenguva dzose backend. Kune Docker uye Kubernetes, chete yeDocker yaunoda kuisa nguva yekumhanya kubva kuNVIDIA, iyo inobvumira maitiro mukati memudziyo kuwana makadhi evhidhiyo mune anotambira. Kubernetes inoda plugin kuitira kuti igone kubata maseva nemakadhi evhidhiyo.
Kusiyana neklassical programming, mune yeML kune zvakawanda zvakasiyana zvinofamba zvinhu muzvivakwa zvinoda kuongororwa nekuyedzwa - semuenzaniso, kodhi yekugadzirisa data, pombi yekudzidzisa modhi uye kugadzira (ona dhairekitori riri pamusoro). Izvo zvakakosha kuti uedze kodhi inobatanidza zvidimbu zvakasiyana zvepombi: kune zvidimbu zvakawanda, uye matambudziko anowanzo simuka pamigero yemamodule.
Iyo AutoML inoshanda sei
AutoML masevhisi anovimbisa kusarudza iyo yakakwana modhi yezvinangwa zvako uye kuidzidzisa. Asi iwe unofanirwa kunzwisisa: data inonyanya kukosha muML, chigumisiro chinoenderana nekugadzirira kwayo. Markup inoitwa nevanhu, iyo yakazara nekukanganisa. Pasina kudzora kwakasimba, mhedzisiro inogona kunge iri marara, uye haisati yave kugona kuita otomatiki maitiro; kuongororwa nenyanzvi - data masayendisiti - inodiwa. Apa ndipo panoputsika AutoML. Asi inogona kubatsira pakusarudza dhizaini - kana iwe watogadzira iyo data uye uchida kumhanyisa nhevedzano yekuyedza kuti uwane yakanakisa modhi.
Nzira yekupinda sei muchina kudzidza
Nzira iri nyore yekupinda muML ndeye kana iwe ukagadzira muPython, iyo inoshandiswa mune ese akadzama ekudzidza masisitimu (uye akajairwa masisitimu). Mutauro uyu unosungirwa kuita basa iri. C ++ inoshandiswa kune mamwe mabasa ekuona komputa, semuenzaniso, mukudzora masisitimu emotokari dzinozvityaira. JavaScript neShell - yekuona uye zvinhu zvinoshamisa sekumhanyisa neuron mubrowser. Java neScala zvinoshandiswa kana uchishanda neBig Data uye pakudzidza muchina. R naJulia vanodiwa nevanhu vanodzidza nhamba dzemasvomhu.
Nzira iri nyore yekuwana ruzivo rwekutanga iri paKaggle; kutora chikamu mune imwe yemakwikwi epuratifomu kunopa rinopfuura gore rekudzidza dzidziso. Papuratifomu iyi unogona kutora kodhi yemumwe munhu yakatumirwa uye yekutaura uye edza kuivandudza, kukwidziridza kune zvinangwa zvako. Bhonasi - yako Kaggle chinzvimbo inokanganisa muhoro wako.
Imwe sarudzo ndeyekujoinha timu yeML semugadziri webackend. Kune akawanda emuchina ekudzidza ekutanga kwaunogona kuwana ruzivo nekubatsira vaunoshanda navo kugadzirisa matambudziko avo. Chekupedzisira, unogona kujoinha imwe yenzvimbo dzesainzi data - Vhura Data Sainzi (ods.ai) nevamwe.
Mukurukuri akaisa rumwe ruzivo nezvemusoro wenyaya pane chinongedzo
"Quadrupel" - sevhisi yakanangwa zviziviso zveiyo portal "State Services"
Evgeny Smirnov
Mukurukuri akatevera aiva mukuru we-e-government infrastructure development department, Evgeny Smirnov, uyo akataura nezve Quadruple. Iri ibasa rakanangwa rekuzivisa yeGosuslugi portal (gosuslugi.ru), iyo inonyanya kushanyirwa hurumende sosi paRunet. Vateereri vezuva nezuva vane mamiriyoni e2,6, pamwe chete kune mamiriyoni makumi mapfumbamwe evashandisi vakanyoreswa panzvimbo ino, iyo 90 mamiriyoni anosimbiswa. Iyo mutoro pane portal API ndeye 60 zviuru RPS.
Tekinoroji dzinoshandiswa kumashure kweState Services
"Quadrupel" ibasa rakanangana nekuzivisa, nerubatsiro rwekuti mushandisi anogamuchira kupihwa kwesevhisi panguva yakakodzera zvakanyanya kwaari nekumisikidza yakakosha yekuzivisa mitemo. Izvo zvinonyanya kudiwa pakugadzira sevhisi zvaive zvigadziriso zvinochinjika uye nguva yakakwana yekutumira mameseji.
Quadrupel inoshanda sei?
Mufananidzo uri pamusoro unoratidza imwe yemitemo yekushanda kweQuadrupel uchishandisa muenzaniso wemamiriro ezvinhu ane chido chekutsiva rezinesi rekutyaira. Chekutanga, sevhisi inotarisa vashandisi vane zuva rekupera basa rinopera mumwedzi. Ivo vanoratidzwa banner ine chipo chekugamuchira iyo yakakodzera sevhisi uye meseji inotumirwa neemail. Kune avo vashandisi vane nguva yekupedzisira yatopera, banner uye email inoshanduka. Mushure mekuchinjana kwakabudirira kwekodzero, mushandisi anogamuchira zvimwe zviziviso - nechikumbiro chekuvandudza iyo data mukuzivikanwa.
Kubva pakuona kwehunyanzvi, aya magwaro egroovy umo iyo code yakanyorwa. Iyo yekupinza idata, yakabuda ichokwadi / nhema, inofananidzwa / haina kuenderana. Kune mitemo inopfuura makumi mashanu muhuwandu - kubva pakusarudza zuva rekuzvarwa kwemushandisi (zuva razvino rakaenzana nezuva rekuzvarwa kwemushandisi) kusvika kumamiriro akaoma. Zuva nezuva, mitemo iyi inozivisa machisi angangoita miriyoni β vanhu vanoda kuziviswa.
Quadrupel yekuzivisa nzira
Pasi pehodhi yeQuadrupel pane dhatabhesi inochengeterwa data remushandisi, uye matatu maapplication:
- Mushandi yakagadzirirwa kuvandudza data.
- Zorora API inotora uye inoendesa iwo mabhena kune iyo portal uye mobile application.
- Muparidzi inotanga basa rekuverenga patsva mabhena kana mameseji akawanda.
Kuti uvandudze data, iyo backend inofambiswa nechiitiko. Nzvimbo mbiri - kuzorora kana JMS. Kune zvakawanda zviitiko; usati wachengeta uye kugadzirisa, iwo akaunganidzwa kuti asaite zvikumbiro zvisina basa. Iyo dhatabhesi pachayo, tafura inochengeterwa data, inoratidzika seyakakosha kukosha chitoro - kiyi yemushandisi uye kukosha kwayo: mireza inoratidza kuvepo kana kusavapo kwemagwaro akakodzera, nguva yavo yechokwadi, yakaunganidzwa manhamba pane kurongeka kwesevhisi. mushandisi uyu, zvichingodaro.
Mushure mekuchengetedza iyo data, basa rinoiswa muJMS kuitira kuti mabhenji adzokororwe nekukurumidza - izvi zvinofanirwa kuratidzwa pakarepo pawebhu. Iyo sisitimu inotanga husiku: mabasa anokandwa muJMS panguva dzevashandisi, zvichienderana nekuti iyo mitemo inoda kuverengerwa zvakare. Izvi zvinotorwa nema processors anobatanidzwa mukuverengazve. Tevere, mhinduro dzekugadzirisa dzinoenda kune inotevera mutsara, iyo inochengetedza mabhana mudhatabhesi kana kutumira mabasa ekuzivisa mushandisi kubasa. Maitiro acho anotora 5-7 maawa, ari nyore scalable nekuda kwekuti iwe unogona kugara uchiwedzera vanobata kana kusimudza zviitiko nevatambi vatsva.
Iyo sevhisi inoshanda chaizvo. Asi huwandu hwe data huri kukura sezvo paine vashandisi vakawanda. Izvi zvinotungamira mukuwedzera kwemutoro pane dhatabhesi - kunyangwe uchifunga nezve iyo Rest API inotarisa replica. Yechipiri pfungwa ndeyeJMS, iyo, sezvazvakazoitika, haina kunyatsokodzera nekuda kwekunyanya kushandiswa kwekuyeuka. Pane njodzi huru yekufashukira kwemutsetse zvichiita kuti JMS iparare uye kugadzirisa kumira. Hazvibviri kusimudza JMS mushure meizvi pasina kubvisa matanda.
Yakarongwa kugadzirisa matambudziko uchishandisa sharding, iyo inobvumira kuenzanisa mutoro pane database. Kune zvakare zvirongwa zvekushandura dhata yekuchengetedza hurongwa, uye shandura JMS kuenda kuKafka - imwe mhinduro yekutadza-inoshivirira inogadzirisa matambudziko ekurangarira.
Backend-as-a-Service Vs. Serverless
Kubva kuruboshwe kuenda kurudyi: Alexander Borgart, Andrey Tomilenko, Nikolay Markov, Ara Israelyan
Backend sevhisi kana Serverless mhinduro? Vatori vechikamu munhaurirano yenyaya iyi inotsikirira patafura yakatenderera vaive:
- Ara Israelyan, CTO CTO uye muvambi we Scorocode.
- Nikolay Markov, Senior Data Engineer kuAligned Research Group.
- Andrey Tomilenko, mukuru weRUVDS Development department.
Hurukuro iyi yakagadziridzwa nemukuru wekuvandudza Alexander Borgart. Tinopa nhaurirano umo vateereri vakapindawo mushanduro yakapfupikiswa.
- Chii chinonzi Serverless mukunzwisisa kwako?
Andrei: Iyi imhando yekombuta - basa reLambda iro rinofanirwa kugadzirisa data kuitira kuti mhedzisiro inoenderana chete nedata. Izwi iri rakauya kubva kuGoogle kana kubva kuAmazon uye neAWS Lambda sevhisi. Zviri nyore kuti mupi aite basa rakadaro nekugovera dziva rekugona kwaro. Vashandisi vakasiyana vanogona kuverengerwa vakazvimirira pamaseva akafanana.
Nikolai: Kuzvitaura zviri nyore, tiri kuendesa chimwe chikamu cheIT yedu zvivakwa uye bhizinesi pfungwa kune gore, kune kunze.
Macaw: Kune chikamu chevagadziri - kuedza kwakanaka kuchengetedza zviwanikwa, kune chikamu chevatengesi - kuwana mari yakawanda.
- Ko Serverless yakafanana ne microservices?
Nikolai: Kwete, Serverless yakawanda yesangano rezvivakwa. A microservice chikamu cheatomu cheimwe pfungwa. Serverless inzira, kwete "chinhu chakaparadzana."
Macaw: A Serverless basa rinogona kuiswa mukati me microservice, asi izvi hazvichazove Serverless, zvicharega kuva Lambda basa. MuServerless, basa rinongotanga kushanda nguva yarinenge rakumbirwa.
Andrei: Vanosiyana muhupenyu hwavo. Takatangisa basa reLambda tikarikanganwa. Yakashanda kwemasekonzi akati wandei, uye mutengi anotevera anogona kugadzirisa chikumbiro chake pane mumwe muchina wenyama.
β Ndeapi zviyero zviri nani?
Macaw: Kana uchiyera zvakachinjika, Lambda mashandiro anoita zvakafanana nemamicroservices.
Nikolai: Chero huwandu hwema replicas aunoisa, pachave neakawanda awo; Serverless haina matambudziko nekuyera. Ini ndakagadzira replica yakaiswa muKubernetes, ndakatangisa makumi maviri zviitiko "kumwe", uye makumi maviri asingazivikanwe malink akadzoserwa kwauri. Pamberi!
- Zvinoita here kunyora backend paServerless?
Andrei: Zvinoreva, asi hazvina musoro. Lambda mabasa anotsamira pane imwechete repository - isu tinofanirwa kuve nechokwadi. Semuyenzaniso, kana mushandisi akaita chimwe kutengeserana, ipapo nguva inotevera yaanozobata anofanirwa kuona: kutengeserana kwaitwa, mari yakaverengerwa. Ese maLambda mabasa anovhara pane iyi runhare. Muchokwadi, boka reServerless mabasa rinoshanduka kuita sevhisi imwechete ine imwe bhodhoro yekuwana nzvimbo kune dhatabhesi.
- Mune mamiriro api zvine musoro kushandisa serverless dhizaini?
Andrei: Mabasa asingade kugovaniswa kuchengetedza - migodhi yakafanana, blockchain. Kwaunoda kuita zvakawanda kuverenga. Kana iwe uine simba rakawanda rekombuta, saka unogona kutsanangura basa se "kuverenga hashi yechimwe chinhu ipapo ..." Asi iwe unogona kugadzirisa dambudziko nekuchengetedza data nekutora, semuenzaniso, Lambda mabasa kubva kuAmazon uye yavo yakagoverwa kuchengetedza. . Uye zvinoitika kuti uri kunyora sevhisi yenguva dzose. Lambda mabasa achawana kuchengetedza uye nekupa imwe mhando yekupindura kumushandisi.
Nikolai: Midziyo inomhanya muServerless yakanyanya kushomeka muzviwanikwa. Pane ndangariro shoma uye zvimwe zvose. Asi kana zvivakwa zvako zvese zvakaiswa zvachose pane rimwe gore - Google, Amazon - uye uine chibvumirano chisingaperi navo, pane bhajeti reizvi zvese, saka kune mamwe mabasa unogona kushandisa Serverless midziyo. Izvo zvinodikanwa kuve mukati mezvivakwa izvi, nekuti zvese zvakagadzirirwa kushandiswa mune yakatarwa nharaunda. Ndokunge, kana iwe wakagadzirira kusungira zvese kune gore rekuvaka, unogona kuyedza. Chakanakira ndechekuti haufanirwe kutonga iyi masisitimu.
Macaw: Icho chokwadi chekuti Serverless haidi kuti iwe utore Kubernetes, Docker, kuisa Kafka, zvichingodaro kuzvinyengedza. Iwo maAmazon neGoogle ari kuisa izvi. Chimwe chinhu ndechekuti une SLA. Iwe unogona zvakare kuburitsa zvese kunze pane kuzvinyora iwe pachako.
Andrei: Serverless pachayo haina kudhura, asi iwe unofanirwa kubhadhara yakawanda kune mamwe maAmazon masevhisi - semuenzaniso, database. Vanhu vakatovamhan'arira nekuti vakachaja mari inopenga pagedhi reAPI.
Macaw: Kana tikataura nezve mari, saka iwe unofanirwa kufunga nezve iyi pfungwa: iwe uchafanirwa kushandura iyo yese nzira yekuvandudza mukambani 180 madhigirii kuitira kuti uendese iyo kodhi yese kuServerless. Izvi zvinotora nguva yakawanda uye mari.
- Pane dzimwe nzira dzakakodzera dzekubhadhara Serverless kubva kuAmazon neGoogle?
Nikolai: MuKubernetes, unotanga imwe mhando yebasa, inomhanya uye inofa - iyi haina Serverless kubva pakuona kwekuvaka. Kana iwe uchida kugadzira inonakidza bhizinesi logic ine mitsetse uye dhatabhesi, saka iwe unofanirwa kufunga zvishoma nezvazvo. Izvi zvese zvinogona kugadziriswa pasina kusiya Kubernetes. Ini handingazvinetse nekukwevera kunze kwekuwedzera kuita.
- Zvakakosha sei kutarisa zviri kuitika muServerless?
Macaw: Zvinoenderana nehurongwa hwekuvaka uye bhizinesi zvinodiwa. Chaizvoizvo, mupi anofanira kupa chirevo chinozobatsira timu ye devops kunzwisisa matambudziko angangoitika.
Nikolai: Amazon ine CloudWatch, uko matanda ese anofambiswa, kusanganisira aya anobva kuLambda. Batanidza kuendesa mberi uye shandisa chishandiso chakasiyana chekutarisa, kunyevera, zvichingodaro. Unogona kuisa vamiririri mumidziyo yaunotanga.
- Ngatitaurei muchidimbu.
Andrei: Kufunga nezve Lambda mabasa kunobatsira. Kana iwe ukagadzira sevhisi iwe pachako - kwete microservice, asi iyo inonyora chikumbiro, inowana iyo dhatabhesi uye inotumira mhinduro - iyo Lambda basa rinogadzirisa akati wandei matambudziko: nemultithreading, scalability, zvichingodaro. Kana pfungwa dzako dzakavakwa nenzira iyi, saka mune ramangwana uchakwanisa kuendesa aya maLambdas kune microservices kana kushandisa wechitatu-bato masevhisi seAmazon. Iyo tekinoroji inobatsira, iyo pfungwa inonakidza. Kuti zvinoruramiswa sei kune bhizinesi ichiri mubvunzo wakavhurika.
Nikolay: Serverless inoshandiswa zviri nani kumabasa ekushanda pane kuverenga mamwe bhizinesi pfungwa. Ndinogara ndichifunga nezvayo sekugadzirisa chiitiko. Kana iwe unayo muAmazon, kana iwe uri muKubernetes, hongu. Zvikasadaro, iwe uchafanirwa kuisa mune kwakawanda kuedza kuti utore Serverless uye ushande wega. Izvo zvinodikanwa kutarisa pane chaiyo bhizinesi kesi. Semuenzaniso, rimwe remabasa angu ikozvino nderokuti: kana mafaera aonekwa pane diski mune imwe fomati, ini ndinofanira kuaisa kuKafka. Ndinogona kushandisa WatchDog kana Lambda. Kubva pakuona kunonzwisisika, zvose zvingasarudzwa zvakakodzera, asi maererano nekushandiswa, Serverless inonyanya kuoma, uye ini ndinosarudza nzira iri nyore, pasina Lambda.
Macaw: Serverless inonakidza, inoshanda, uye yakanyanya hunyanzvi zano rakanaka. Nenguva isipi, tekinoroji ichasvika padanho rekuti chero basa richatangwa mune isingasviki 100 milliseconds. Zvadaro, musimboti, hapazovi nemubvunzo wekuti nguva yekumirira yakakosha kumushandisi. Panguva imwecheteyo, kushandiswa kweServerless, sezvakataurwa nevamwe vashandi, zvakakwana zvinoenderana nechinetso chebhizimisi.
Tinotenda vatsigiri vedu vakatibatsira zvakanyanya:
- IT nzvimbo yemusangano «»kunzvimbo yemusangano.
- Karenda yezviitiko zveIT uye kubudiswa"Β» Kuti uwane ruzivo uye ruzivo.
- Β«"zvezvipo.
- zvekugadzira pamwe chete.
- "Asociation for Electronic Communications" nokuda kwekubatanidzwa uye ruzivo.
- Main mubatsiri - kune vese!
Source: www.habr.com