Msonkhano wa Habr si nkhani yoyamba. M'mbuyomu, tidachita zochitika zazikulu za Toaster kwa anthu 300-400, koma tsopano tidaganiza kuti misonkhano yaying'ono ingakhale yoyenera, momwe mungakhazikitsire, mwachitsanzo, mu ndemanga. Msonkhano woyamba wamtunduwu udachitika mu Julayi ndipo adadzipereka kuti athandizire chitukuko. Ophunzirawo adamvetsera malipoti okhudza kusintha kuchokera kumbuyo kupita ku ML komanso za mapangidwe a ntchito ya Quadrupel pa portal ya State Services, komanso adatenga nawo gawo pa tebulo lozungulira loperekedwa kwa Serverless. Kwa iwo omwe sanathe kukhala nawo pamwambowu pamasom'pamaso, mu positi iyi tikukuuzani zinthu zosangalatsa kwambiri.
Kuchokera ku chitukuko chakumbuyo mpaka kuphunzira makina
Kodi mainjiniya a data amachita chiyani ku ML? Kodi ntchito za wopanga backend ndi injiniya wa ML ndizofanana bwanji komanso zosiyana? Kodi muyenera kuchita chiyani kuti musinthe ntchito yanu yoyamba kukhala yachiwiri? Izi zidanenedwa ndi Alexander Parinov, yemwe adapita kukaphunzira makina pambuyo pa zaka 10 za ntchito yakumbuyo.
Alexander Parinov
Masiku ano Alexander amagwira ntchito ngati womanga makina owonera makompyuta ku X5 Retail Group ndipo amathandizira mapulojekiti a Open Source okhudzana ndi masomphenya apakompyuta komanso kuphunzira mozama (github.com/creafz). Maluso ake amatsimikiziridwa ndi kutenga nawo mbali mu 100 yapamwamba ya dziko la Kaggle Master (kaggle.com/creafz), nsanja yotchuka kwambiri ya mpikisano wophunzirira makina.
Chifukwa chiyani kusintha kuphunzira makina
Chaka ndi theka chapitacho, Jeff Dean, wamkulu wa Google Brain, kafukufuku wozama wa Google wogwiritsa ntchito nzeru zopangapanga, adafotokoza momwe mizere ya theka la miliyoni mu Google Translate idasinthidwa ndi netiweki ya Tensor Flow neural yokhala ndi mizere ya 500 yokha. Pambuyo pophunzitsa maukonde, khalidwe la deta linakula ndipo zomangamanga zinakhala zosavuta. Zingawoneke kuti ili ndi tsogolo lathu lowala: sitiyeneranso kulemba code, ndikwanira kupanga ma neuroni ndikuwadzaza ndi deta. Koma muzochita zonse zimakhala zovuta kwambiri.
ML Infrastructure ku Google
Neural network ndi gawo laling'ono chabe la zomangamanga (malo ang'onoang'ono akuda pachithunzi pamwambapa). Machitidwe ambiri othandizira amafunikira kuti alandire deta, kuikonza, kuisunga, kuyang'ana khalidwe, ndi zina zotero, timafunikira zipangizo zophunzitsira, kutumiza makina ophunzirira makina popanga, ndikuyesa code iyi. Ntchito zonsezi ndizofanana ndendende ndi zomwe opanga backend amachita.
Njira yophunzirira makina
Kodi pali kusiyana kotani pakati pa ML ndi backend?
M'mapulogalamu apamwamba, timalemba code ndipo izi zimayang'anira khalidwe la pulogalamuyo. Mu ML, tili ndi kachidindo kakang'ono kachitsanzo ndi deta zambiri zomwe timaponyera pa chitsanzo. Deta mu ML ndi yofunika kwambiri: chitsanzo chomwecho chophunzitsidwa pazinthu zosiyanasiyana chikhoza kusonyeza zotsatira zosiyana kwambiri. Vuto ndilakuti zambiri zimabalalika nthawi zonse ndikusungidwa m'makina osiyanasiyana (zosungirako zaubale, ma database a NoSQL, zipika, mafayilo).
Kusintha kwa data
ML imafuna kumasulira osati kachidindo kokha, monga mwachitukuko chachikale, komanso deta: m'pofunika kumvetsetsa bwino zomwe chitsanzocho chinaphunzitsidwa. Kuti muchite izi, mutha kugwiritsa ntchito laibulale yotchuka ya Data Science Version Control (dvc.org).
Kuyika kwa data
Ntchito yotsatira ndikulemba zolemba. Mwachitsanzo, chongani zinthu zonse pachithunzichi kapena nenani kuti ndi za gulu liti. Izi zimachitika ndi mautumiki apadera monga Yandex.Toloka, ntchito yomwe imakhala yosavuta kwambiri ndi kukhalapo kwa API. Zovuta zimadza chifukwa cha "chinthu chaumunthu": mutha kusintha kuchuluka kwa data ndikuchepetsa zolakwika pang'ono popereka ntchito yomweyo kwa ochita angapo.
Kuwoneka mu Tensor Board
Kudula zoyeserera ndikofunikira kuti mufananize zotsatira ndikusankha mtundu wabwino kwambiri potengera ma metrics ena. Pali zida zazikulu zowonera - mwachitsanzo, Tensor Board. Koma palibe njira zabwino zosungira zoyeserera. Makampani ang'onoang'ono nthawi zambiri amachita ndi Excel spreadsheet, pomwe akuluakulu amagwiritsa ntchito nsanja zapadera posungira zotsatira mu database.
Pali nsanja zambiri zophunzirira makina, koma palibe imodzi yomwe imakwaniritsa 70% ya zosowa
Vuto loyamba lomwe munthu ayenera kukumana nalo poika chitsanzo chophunzitsidwa bwino pakupanga likugwirizana ndi chida chomwe amachikonda kwambiri asayansi a data - Jupyter Notebook. Palibe modularity mmenemo, ndiko kuti, zotuluka ndi "nsalu ya phazi" ya code yomwe siinagawidwe mu zidutswa zomveka - ma modules. Chilichonse chimasakanizidwa: makalasi, ntchito, masinthidwe, ndi zina. Khodi iyi ndi yovuta kuyisintha ndikuyesa.
Kodi kuthana ndi izi? Mutha kusiya ntchito, monga Netflix, ndikupanga nsanja yanu yomwe imakulolani kuti mutsegule ma laputopuwa mwachindunji pakupanga, kusamutsa deta kwa iwo monga zolowera ndikupeza zotsatira. Mutha kukakamiza opanga omwe akuyendetsa mtunduwu kuti alembenso kachidindo nthawi zonse, kuiphwanya kukhala ma module. Koma ndi njira iyi n'zosavuta kulakwitsa, ndipo chitsanzo sichingagwire ntchito monga momwe tafunira. Chifukwa chake, njira yabwino ndikuletsa kugwiritsa ntchito Jupyter Notebook pama code code. Ngati, ndithudi, asayansi a data amavomereza izi.
Chitsanzo ngati bokosi lakuda
Njira yosavuta yopezera chitsanzo pakupanga ndikuigwiritsa ntchito ngati bokosi lakuda. Muli ndi mtundu wina wa kalasi yachitsanzo, munapatsidwa zolemera zachitsanzo (magawo a ma neuroni a netiweki yophunzitsidwa), ndipo ngati mutayambitsa kalasi iyi (itanani njira yolosera, idyetseni chithunzi), mudzapeza zina. kulosera ngati zotsatira. Zomwe zimachitika mkatimo zilibe kanthu.
Phatikizani ndondomeko ya seva ndi chitsanzo
Mutha kukwezanso njira ina yosiyana ndikuitumiza pamzere wa RPC (ndi zithunzi kapena deta ina. Pazotulutsa tidzalandira zolosera.
Chitsanzo chogwiritsira ntchito chitsanzo mu Flask:
@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)
Vuto ndi njira iyi ndi kuchepa kwa magwiridwe antchito. Tinene kuti tili ndi Phyton code yolembedwa ndi asayansi a data yomwe imachedwa, ndipo tikufuna kufinya magwiridwe antchito. Kuti muchite izi, mutha kugwiritsa ntchito zida zomwe zimasinthira kachidindo kukhala komweko kapena kusinthira kukhala chimango china chopangidwira kupanga. Pali zida zotere pa chimango chilichonse, koma palibe zabwino; muyenera kuziwonjezera nokha.
Zomangamanga mu ML ndizofanana ndi kumbuyo kwanthawi zonse. Pali Docker ndi Kubernetes, kwa Docker kokha muyenera kukhazikitsa nthawi yothamanga kuchokera ku NVIDIA, yomwe imalola njira mkati mwa chidebe kuti mupeze makhadi avidiyo omwe ali nawo. Kubernetes amafunikira pulogalamu yowonjezera kuti athe kuyang'anira ma seva okhala ndi makhadi avidiyo.
Mosiyana ndi mapulogalamu akale, pankhani ya ML pali zinthu zambiri zosunthika pazomangamanga zomwe ziyenera kufufuzidwa ndikuyesedwa - mwachitsanzo, ma code processing data, payipi yophunzitsira yachitsanzo ndi kupanga (onani chithunzi pamwambapa). Ndikofunikira kuyesa kachidindo komwe kumagwirizanitsa mapaipi osiyanasiyana: pali zidutswa zambiri, ndipo mavuto amadza nthawi zambiri pamalire a module.
Momwe AutoML imagwirira ntchito
Ntchito za AutoML zimalonjeza kusankha mtundu woyenera pazolinga zanu ndikuphunzitsa. Koma muyenera kumvetsetsa: deta ndi yofunika kwambiri mu ML, zotsatira zake zimadalira kukonzekera kwake. Markup imachitidwa ndi anthu, omwe ali odzaza ndi zolakwika. Popanda kuwongolera mosamalitsa, zotsatira zake zitha kukhala zinyalala, ndipo sikutheka kupanga makinawo; kutsimikiziridwa ndi akatswiri - asayansi a data - ndikofunikira. Apa ndipamene AutoML imasweka. Koma zingakhale zothandiza posankha zomangamanga - pamene mwakonzekera kale deta ndipo mukufuna kuyendetsa mayesero angapo kuti mupeze chitsanzo chabwino.
Momwe mungalowe mu maphunziro a makina
Njira yosavuta yolowera mu ML ndikukhala mu Python, yomwe imagwiritsidwa ntchito pamaphunziro akuzama (ndi machitidwe okhazikika). Chilankhulochi ndi chovomerezeka pazochitika izi. C ++ imagwiritsidwa ntchito pazinthu zina zamakompyuta, mwachitsanzo, pakuwongolera magalimoto odziyendetsa okha. JavaScript ndi Shell - zowonera ndi zinthu zachilendo monga kuyendetsa neuron mu msakatuli. Java ndi Scala amagwiritsidwa ntchito pogwira ntchito ndi Big Data komanso kuphunzira pamakina. R ndi Julia amakondedwa ndi anthu amene amaphunzira masamu.
Njira yabwino kwambiri yopezera chidziwitso choyambira ndi pa Kaggle; kutenga nawo mbali mumpikisano umodzi wapapulatifomu kumapereka kupitilira chaka chimodzi chophunzirira chiphunzitso. Pa nsanja mukhoza kutenga wina anaika ndi ndemanga kachidindo ndi kuyesa kusintha, kukhathamiritsa ndi zolinga zanu. Bonasi - udindo wanu wa Kaggle umakhudza malipiro anu.
Njira ina ndikulowa nawo gulu la ML ngati woyambitsa backend. Pali zoyambira zambiri zophunzirira makina komwe mungapeze luso pothandiza anzanu kuthetsa mavuto awo. Pomaliza, mutha kujowina gulu limodzi la asayansi a data - Open Data Science (ods.ai) ndi ena.
Wokamba nkhaniyo adatumiza zowonjezera pamutuwu pa ulalo
"Quadrupel" - ntchito ya zidziwitso zomwe zikugwirizana ndi portal "State Services"
Evgeny Smirnov
Wokamba nkhani wotsatira anali mtsogoleri wa dipatimenti ya chitukuko cha e-government, Evgeny Smirnov, yemwe analankhula za Quadruple. Uwu ndi ntchito yodziwitsira zidziwitso za Gosuslugi portal (gosuslugi.ru), zomwe zidayendera kwambiri boma pa Runet. Omvera tsiku ndi tsiku ndi 2,6 miliyoni, palimodzi pali 90 miliyoni olembetsa olembetsa patsamba, omwe 60 miliyoni amatsimikiziridwa. Katundu pa portal API ndi 30 zikwi RPS.
Matekinoloje omwe amagwiritsidwa ntchito kumbuyo kwa State Services
"Quadrupel" ndi ntchito yazidziwitso yomwe imayang'aniridwa, mothandizidwa ndi zomwe wogwiritsa ntchito amalandila chithandizo panthawi yoyenera kwambiri kwa iye mwa kukhazikitsa malamulo apadera azidziwitso. Zofunikira zazikulu popanga ntchitoyo zinali zosinthika komanso nthawi yokwanira yotumizira makalata.
Kodi Quadrupel imagwira ntchito bwanji?
Chithunzi pamwambapa chikuwonetsa imodzi mwamalamulo ogwirira ntchito a Quadrupel pogwiritsa ntchito chitsanzo cha vuto lomwe likufunika kusintha chiphaso choyendetsa. Choyamba, ntchitoyi imayang'ana ogwiritsa ntchito omwe tsiku lawo lotha ntchito limatha mwezi umodzi. Amawonetsedwa chikwangwani chokhala ndi mwayi woti alandire chithandizo choyenera ndipo uthenga umatumizidwa ndi imelo. Kwa ogwiritsa ntchito omwe nthawi yawo yomaliza yatha kale, banner ndi imelo zimasintha. Pambuyo pakusinthana bwino kwaufulu, wogwiritsa ntchito amalandira zidziwitso zina - ndi lingaliro losinthira deta muzodziwika.
Kuchokera pamalingaliro aukadaulo, awa ndi ma groovy scripts momwe code imalembedwera. Zomwe zalowetsedwa ndi deta, zotulukapo ndi zoona / zabodza, zofananira / sizinafanane. Pali malamulo opitilira 50 onse - kuyambira pakuzindikira tsiku lobadwa la wogwiritsa ntchito (tsiku lomwe lili pano ndi lofanana ndi tsiku lobadwa la wogwiritsa ntchito) mpaka zovuta. Tsiku lililonse, malamulowa amazindikiritsa machesi pafupifupi miliyoni miliyoni βanthu ofunikira kudziwitsidwa.
Makanema azidziwitso a Quadrupel
Pansi pa Quadrupel pali nkhokwe yomwe deta ya ogwiritsa ntchito imasungidwa, ndi ntchito zitatu:
- Wogwira ntchito cholinga chosinthira deta.
- API Yotsalira imatenga ndikupereka zikwangwani zokha ku portal ndi pulogalamu yam'manja.
- Scheduler imayambitsa ntchito yowerengeranso zikwangwani kapena kutumiza makalata ambiri.
Kuti musinthe deta, backend imayendetsedwa ndi zochitika. Mawonekedwe awiri - kupuma kapena JMS. Pali zochitika zambiri; musanayambe kupulumutsa ndi kukonza, amaphatikizidwa kuti asapange zopempha zosafunikira. Dongosolo lokhalokha, tebulo lomwe deta imasungidwa, limawoneka ngati sitolo yamtengo wapatali - fungulo la wogwiritsa ntchito komanso mtengo wake: mbendera zomwe zikuwonetsa kukhalapo kapena kusakhalapo kwa zikalata zoyenera, nthawi yovomerezeka, ziwerengero zophatikiza pa dongosolo la ntchito ndi wosuta uyu, ndi zina zotero.
Pambuyo posunga deta, ntchito imayikidwa mu JMS kuti zikwangwani ziwerengedwenso nthawi yomweyo - izi ziyenera kuwonetsedwa nthawi yomweyo pa intaneti. Dongosolo limayamba usiku: ntchito zimaponyedwa mu JMS pakapita nthawi, malinga ndi zomwe malamulowo amayenera kuwerengedwanso. Izi zimatengedwa ndi ma processor omwe akuphatikizidwa pakuwerengeranso. Chotsatira, zotsatira zokonzekera zimapita pamzere wotsatira, womwe umasunga zikwangwani mu database kapena kutumiza ntchito zodziwitsa ogwiritsa ntchito ku ntchitoyo. Njirayi imatenga maola 5-7, imatha kuchulukirachulukira chifukwa mutha kuwonjezera othandizira kapena kukweza zochitika ndi othandizira atsopano.
Ntchitoyi imagwira ntchito bwino. Koma kuchuluka kwa deta kukukulirakulira chifukwa pali ogwiritsa ntchito ambiri. Izi zimabweretsa kuwonjezeka kwa katundu pa database - ngakhale poganizira kuti Rest API imayang'ana chithunzicho. Mfundo yachiwiri ndi JMS, yomwe, monga momwe zinakhalira, sizoyenera kwambiri chifukwa chakugwiritsa ntchito kukumbukira kwambiri. Pali chiwopsezo chachikulu chakusefukira kwa mizere kupangitsa kuti JMS iwonongeke ndikuyimitsa. Ndizosatheka kukweza JMS pambuyo pa izi popanda kuchotsa zipika.
Zakonzedwa kuthetsa mavuto pogwiritsa ntchito sharding, zomwe zidzalola kugwirizanitsa katundu pa database. Palinso mapulani osintha ndondomeko yosungiramo deta, ndikusintha JMS kukhala Kafka - njira yothetsera vuto lomwe lidzathetse mavuto a kukumbukira.
Backend-as-a-Service vs. Wopanda seva
Kuchokera kumanzere kupita kumanja: Alexander Borgart, Andrey Tomilenko, Nikolay Markov, Ara Israelyan
Backend ngati ntchito kapena Serverless yankho? Otenga nawo gawo pazokambitsirana zankhani yomwe ikufunikayi pa tebulo lozungulira anali:
- Ara Israelyan, CTO CTO ndi woyambitsa Scorocode.
- Nikolay Markov, Senior Data Engineer at Aligned Research Group.
- Andrey Tomilenko, wamkulu wa dipatimenti yachitukuko ya RUVDS.
Zokambiranazo zidayendetsedwa ndi wopanga wamkulu Alexander Borgart. Timapereka makangano omwe omvera nawonso adatenga nawo gawo mu chidule chachidule.
- Kodi Serverless mu kumvetsetsa kwanu ndi chiyani?
Andrei: Ichi ndi chitsanzo cha makompyuta - ntchito ya Lambda yomwe imayenera kukonza deta kuti zotsatira zake zimadalira deta yokha. Mawuwa adachokera ku Google kapena kuchokera ku Amazon ndi ntchito yake ya AWS Lambda. Ndikosavuta kuti wothandizira agwire ntchito yotereyi popereka dziwe la mphamvu zake. Ogwiritsa ntchito osiyanasiyana amatha kudziwerengera okha pa ma seva omwewo.
Nikolai: Kunena mwachidule, tikusamutsa gawo lina lazinthu zathu za IT ndi malingaliro abizinesi kupita kumtambo, kupita ku kunja.
Ara: Kumbali ya omanga - kuyesa kwabwino kupulumutsa chuma, kwa ogulitsa - kupeza ndalama zambiri.
- Kodi Serverless ndi yofanana ndi ma microservices?
Nikolai: Ayi, Serverless ndi gulu lazomangamanga. Microservice ndi gawo la atomiki lamalingaliro ena. Serverless ndi njira, osati "gulu losiyana."
Ara: Ntchito Yopanda Seva ikhoza kuikidwa mu microservice, koma izi sizidzakhalanso Zopanda Server, zidzasiya kukhala ntchito ya Lambda. Mu Serverless, ntchito imangoyamba kugwira ntchito ikafunsidwa.
Andrei: Amasiyana m'moyo wawo. Tidayambitsa ntchito ya Lambda ndikuyiwala. Zinagwira ntchito kwa masekondi angapo, ndipo kasitomala wotsatira akhoza kukonza pempho lake pamakina ena akuthupi.
- Ndi mamba ati abwinoko?
Ara: Mukakulitsa mopingasa, ntchito za Lambda zimakhala zofanana ndendende ndi ma microservices.
Nikolai: Zirizonse zofananira zomwe mungakhazikitse, zidzakhala zochulukirapo; Serverless ilibe vuto pakukulitsa. Ndidapanga chofanizira ku Kubernetes, ndikuyambitsa zochitika 20 "kwinakwake", ndipo maulalo 20 osadziwika adabwezedwa kwa inu. Patsogolo!
- Kodi n'zotheka kulemba backend pa Serverless?
Andrei: Mwachidziwitso, koma sizomveka. Ntchito za Lambda zidzadalira malo amodzi - tiyenera kutsimikizira. Mwachitsanzo, ngati wogwiritsa ntchitoyo adachitapo kanthu, ndiye kuti nthawi ina akadzakumananso ayenera kuwona: ntchitoyo yachitika, ndalamazo zawerengedwa. Ntchito zonse za Lambda zidzatsekereza kuyimba uku. M'malo mwake, gulu la ntchito zopanda Serverless lidzasanduka ntchito imodzi yokhala ndi malo amodzi ofikira botolo ku database.
- Muzochitika ziti zomwe zimakhala zomveka kugwiritsa ntchito zomangamanga zopanda seva?
Andrei: Ntchito zomwe sizifuna kugawana nawo - migodi yomweyo, blockchain. Kumene muyenera kuwerengera zambiri. Ngati muli ndi mphamvu zambiri zamakompyuta, ndiye kuti mungathe kufotokozera ntchito monga "kuwerengera hashi ya chinachake kumeneko ..." Koma mutha kuthetsa vutoli ndi kusungirako deta potenga, mwachitsanzo, ntchito za Lambda kuchokera ku Amazon ndi kusungidwa kwawo komweko. . Ndipo zikuwoneka kuti mukulemba ntchito yokhazikika. Ntchito za Lambda zidzafikira zosungirako ndikupereka yankho lamtundu wina kwa wogwiritsa ntchito.
Nikolai: Zotengera zomwe zimayenda mu Serverless ndizochepa kwambiri pazinthu. Pali kukumbukira pang'ono ndi china chirichonse. Koma ngati zomangamanga zanu zonse zikugwiritsidwa ntchito pamtambo - Google, Amazon - ndipo muli ndi mgwirizano wokhazikika ndi iwo, pali bajeti ya zonsezi, ndiye kuti ntchito zina mungagwiritse ntchito zotengera za Serverless. Ndikofunikira kukhala mkati mwachitukuko ichi, chifukwa chilichonse chimapangidwa kuti chigwiritsidwe ntchito pamalo enaake. Ndiko kuti, ngati mwakonzeka kumangiriza chirichonse kuzinthu zowonongeka zamtambo, mukhoza kuyesa. Ubwino wake ndikuti simuyenera kuyang'anira izi.
Ara: Mfundo yakuti Serverless sikufuna kuti muyang'ane Kubernetes, Docker, kukhazikitsa Kafka, ndi zina zotero ndikudzinyenga nokha. Amazon ndi Google omwewo akuyika izi. Chinthu china ndi chakuti muli ndi SLA. Mutha kutulutsa chilichonse m'malo mozilemba nokha.
Andrei: Serverless palokha ndi yotsika mtengo, koma muyenera kulipira zambiri pazinthu zina za Amazon - mwachitsanzo, database. Anthu adawasumira kale chifukwa adalipira ndalama zopenga pachipata cha API.
Ara: Ngati tilankhula za ndalama, ndiye kuti muyenera kuganizira mfundo iyi: muyenera kutembenuza njira yonse yachitukuko mu kampani 180 madigiri kuti mutumize code yonse ku Serverless. Izi zidzatenga nthawi ndi ndalama zambiri.
- Kodi pali njira zina zolipirira Serverless kuchokera ku Amazon ndi Google?
Nikolai: Ku Kubernetes, mumayambitsa mtundu wina wa ntchito, imathamanga ndikufa - iyi ndi Serverless ndithu kuchokera kumalo omangamanga. Ngati mukufuna kupanga malingaliro osangalatsa abizinesi okhala ndi mizere ndi nkhokwe, ndiye kuti muyenera kuganiza mozama za izi. Zonsezi zitha kuthetsedwa osasiya Kubernetes. Sindingavutike kukokera kunja kukhazikitsa kowonjezera.
- Ndikofunikira bwanji kuyang'anira zomwe zikuchitika mu Serverless?
Ara: Zimatengera kamangidwe kadongosolo komanso zofunikira zamabizinesi. Kwenikweni, woperekayo ayenera kupereka malipoti omwe angathandize gulu la devops kumvetsetsa zovuta zomwe zingachitike.
Nikolai: Amazon ili ndi CloudWatch, komwe mitengo yonse imaseweredwa, kuphatikiza aku Lambda. Gwirizanitsani kutumiza kwa chipika ndikugwiritsa ntchito chida chosiyana powonera, kuchenjeza, ndi zina zotero. Mutha kuyika othandizira muzotengera zomwe mwayambitsa.
- Tiyeni tifotokoze mwachidule.
Andrei: Kuganizira za ntchito za Lambda ndizothandiza. Ngati mupanga ntchito nokha - osati microservice, koma yomwe imalemba pempho, imapeza deta ndikutumiza yankho - ntchito ya Lambda imathetsa mavuto angapo: ndi multithreading, scalability, ndi zina zotero. Ngati malingaliro anu amangidwa motere, ndiye kuti mtsogolomu mudzatha kusamutsa ma Lambdas ku ma microservices kapena kugwiritsa ntchito ntchito za chipani chachitatu monga Amazon. Zipangizo zamakono ndizothandiza, lingalirolo ndi losangalatsa. Kulungamitsidwa kotani kwa bizinesi likadali funso lotseguka.
Nikolay: Serverless imagwiritsidwa ntchito bwino pantchito zogwirira ntchito kuposa kuwerengera malingaliro abizinesi. Nthawi zonse ndimaganiza ngati kukonza zochitika. Ngati muli nazo ku Amazon, ngati muli ku Kubernetes, inde. Kupanda kutero, mudzayenera kuchita khama kwambiri kuti Serverless ayambe kuyendetsa nokha. Ndikofunikira kuyang'ana nkhani inayake yabizinesi. Mwachitsanzo, imodzi mwa ntchito zanga tsopano ndi: mafayilo akawoneka pa disk mumtundu wina, ndiyenera kuwayika ku Kafka. Nditha kugwiritsa ntchito WatchDog kapena Lambda. Kuchokera pamalingaliro omveka, zosankha zonse ziwiri ndi zoyenera, koma potsata kukhazikitsa, Serverless ndizovuta kwambiri, ndipo ndimakonda njira yosavuta, popanda Lambda.
Ara: Serverless ndi lingaliro losangalatsa, lothandiza, komanso lokongola mwaukadaulo. Posakhalitsa, luso lamakono lidzafika poti ntchito iliyonse idzayambitsidwe m'munsi mwa 100 milliseconds. Ndiye, kwenikweni, sipadzakhala kukayikira ngati nthawi yodikira ndi yofunika kwambiri kwa wogwiritsa ntchito. Panthawi imodzimodziyo, kugwiritsidwa ntchito kwa Serverless, monga momwe anzako anenera kale, kumadalira kwathunthu vuto la bizinesi.
Tikuthokoza othandizira athu omwe atithandiza kwambiri:
- Malo ochitira misonkhano ya IT «»pamalo amsonkhano.
- Kalendala ya zochitika za IT ndi kusindikiza "Β» kuti mudziwe zambiri komanso nkhani.
- Β«"za mphatso.
- za kupanga co-kupanga.
- "Association for Electronic Communications" chifukwa chotengapo mbali komanso chidziwitso.
- Wothandizira wamkulu - kwa onse!
Source: www.habr.com