La Ukuqaphela sekusuke ekubeni yisihloko esibalulekile sobuchwepheshe kwaba yinsika yesu Kunoma iyiphi inhlangano ethembele kusofthiwe—cishe yonke—ukumane “ukuqapha amaseva” noma ukubheka amadeshibhodi ahlukanisiwe akusanele. Izinkampani zidinga ukuqonda ukuthi kwenzekani ngaphakathi kwezinhlelo zazo ngesikhathi sangempela, zixhumanise leyo datha nebhizinisi, futhi zisabele ngokushesha lapho kukhona okungahambi kahle. Futhi, ngaphezu kwakho konke, kumele zenze kanjalo endaweni eqhutshwa kakhulu yisofthiwe. I-AI ye-ejenti, amazinga avulekile, kanye nezakhiwo ezisatshalalisiwe.
Kulesi simo, kusobala ukuthi lo mkhuba ubheke ku- ukubonwa okuvulekile kakhudlwana, okuhlobene kakhulu nemiphumela yebhizinisi kanye nokuzimela okukhuluI-OpenTelemetry isiqala ukusungulwa njengolimi oluvamile lwe-telemetry, i-AI idlulela ngale kokuhlola ukuze ihlanganiswe nesisekelo samapulatifomu okubona, futhi amaqembu e-IPop aguquka abe abahleli bezinhlelo ezihlakaniphile ezithola, zihlaziye, futhi zilungise izinkinga zodwa. Ake sichaze ukuthi lolu shintsho lwenzeka kanjani nokuthi luthinta kanjani ubuchwepheshe, ibhizinisi, ukuphepha, kanye nokuphathwa kwedatha.
Kusukela ekuqapheni okuvamile kuya enkathini yokubonakala
Ukuvela kwesimo sezulu ukuqapha kwendabuko okubheke ekubonweni kwesimanje Kuqala emuva kakhulu. Lapho kuvela amathuluzi okuqala e-APM, njengalawo athandwa yiLew Cirne ngeNew Relic, izindaba ezinkulu kwakuwukukwazi ukubona ngokuningiliziwe ukuthi ikhodi yohlelo lokusebenza lwe-monolithic yenzani esikhungweni sedatha esiphethwe yinkampani. Lokho kwakuyinguquko: okokuqala ngqa, amaqembu ayengabona ukusebenza kwezinhlelo zawo zokukhiqiza nge-granularity enhle kakhulu.
Ngokufika kwe ukubala ngamafu, ama-microservices, izitsha, ukubala okungenaseva, kanye nemikhuba ye-DevOps kanye ne-SREIsimo sashintsha ngokuphelele. Ukushintsha kusuka ezinhlelweni ezihlanganisiwe kuya ezinhlelweni ezisatshalaliswa kwasho ukuthi ukubonakala kwesikhathi akusanele. Isevisi ayisaseyona uhlelo lokusebenza olulodwa, kodwa iyinqwaba yezinsizakalo ezincane zesikhashana, ezihlelwe kumapulatifomu afana ne-Kubernetes, ezisatshalaliswa izikhathi eziningi ngosuku, futhi ezisebenza ezingqalasizinda ezihlanganisiwe nabahlinzeki abaningi bamafu.
Kuleyo ndawo, ukuqapha okuvamile, okugxile kuma-metric achazwe kusengaphambili kanye nezexwayiso ezimile, akuphumelelanga. Ukuqaphela kwethula indlela ehlukile: ukuqoqa nokuxhumanisa izibalo, izingodo, imikhondo, kanye nemicimbi ukuthola isimo sangaphakathi sohlelo kusukela emiphumeleni yalo yangaphandle. Akukhona nje ukwazi ukuthi kukhona okwehlulekayo, kodwa mayelana nokuqonda ukuthi kungani kwenzeke nokuthi kunamuphi umthelela kumsebenzisi kanye nebhizinisi.
Ababhali bathanda Yuri Shkuro Lo mehluko ufingqiwe kahle: ukuqapha kulinganisa lokho okunqunywe kusengaphambili njengokubalulekile, kuyilapho ukuqaphela kukuvumela ukuthi wakhe imibuzo emisha mayelana nohlelo ngaphandle kokulungiselela zonke izinkomba kusengaphambili. Ngamanye amazwi, Ukuqaphela kuguqula idatha ye-telemetry ibe umongo ongasetshenziswa kwentuthuko, imisebenzi kanye nebhizinisi.
Lolu shintsho luqhutshwa futhi yizici ezithile kakhulu: a ingcindezi enonya yokuqamba izinto ezintsha ngokusheshaAmakhasimende afuna kakhulu ukushiya uhlelo lokusebenza ngephutha elincane kakhulu, ububanzi obungenamkhawulo bobuchwepheshe nezinsizakalo eziphethwe, kanye nokukhula ukuzenzekela kwayo yonke i-software lifecycleKonke lokho kuzenzakalelayo kuyisofthiwe engahluleka, futhi idinga ukubonwa kwayo.
Ubunzima, ingozi, namathuluzi amaningi kakhulu: kungani ukuqaphela kubalulekile
Izakhiwo zanamuhla zinezinkinga ezine ezinkulu ezenza ukubonwa cishe kuyimpoqo Uma ufuna ukulawula:
Okokuqala, i- ubunzima buye banda kakhuluIsitsha singaphila imizuzu noma imizuzwana, i-microservice ingashintsha izinguqulo izikhathi eziningana ngosuku, futhi izingxenye ziyanda. Lokho okwakuyi-monolithic application kuba yiqoqo lezinsizakalo ezixhumene. Amaqembu okusebenza azithola ebhekene namakhulu noma izinkulungwane zezinhlangano ezishintsha njalo, eziningi zazo angazange azithuthukise zona.
Ngaphezu kwalokhu ukwanda okucacile kwengoziUkusebenzisa izikhathi eziningi ngosuku kusho ukwethula izinguquko njalo—kanye nokubuyiselwa emuva okungenzeka. Imikhuba ye-Agile kanye nokulethwa okuqhubekayo kunezela amathuluzi engeziwe, amapayipi, kanye nokuzenzakalelayo okudingeka kucatshangelwe. Ikhono lokubona inkinga ngokushesha, ukuhlonza imbangela eyinhloko, nokuyibuyisela emuva noma ukuyilungisa ngemizuzu embalwa akusadingeki kodwa kuyimfuneko.
Ngesikhathi esifanayo, igebe lamakhonoInqwaba yobuchwepheshe inkulu kakhulu kangangokuthi akunakwenzeka ngomuntu oyedwa ukuba abe yingcweti kudathabheyisi, amanethiwekhi, ama-API, ezokuphepha, amakhonteyina, amapulatifomu okuhlanganisa, kanye namathuluzi e-CI/CD. Kudingeka izindlela zokusiza ukuqonda ukuthi konke kuhambisana kanjani, kuncike kulokho, nokuthi kufanele kubhekwe kuphi lapho kukhona okungahambi kahle. Ngaphandle kwalombono oxhumene, isikhathi esichithwayo sigxuma phakathi kwamathuluzi singaba sikhulu kakhulu.
Futhi, ngaphezu kwakho konke, kuvela izinkinga nge "ukusabalala kwamathuluzi" noma amathuluzi amaningiIngqimba ngayinye yesitaki ngokuvamile inesixazululo sayo sokuqapha: esinye sedathabheyisi, esinye sengqalasizinda, esinye sengxenye engaphambili, esinye sezingodo, esinye semikhondo… Ukuhlanganisa idatha phakathi kwazo kuhilela ukushintshashintsha komongo okuqhubekayo, ukusesha ngesandla, kanye nezikhathi ezinde zokuxazulula izigameko. Lokhu kuphambene ngqo nalokho okudingekayo lapho uhlelo lokusebenza lungasebenzi futhi abasebenzisi bekhononda.
Impendulo yakho konke lokhu itholakala ku- ipulatifomu yokubuka ehlanganisiwe eqoqa yonke i-telemetry efanele, iyixhumanise nezinhlangano eziyikhiqizayo, futhi ivumela noma yiliphi iqembu—intuthuko, imisebenzi, ezokuphepha, ibhizinisi—ukuhlola nokusebenzisa leyo datha endaweni eyodwa. Lokhu akubandakanyi nje kuphela izilinganiso zokusebenza kodwa nemicimbi yebhizinisi kanye nezimpawu eziveza umthelela wezomnotho wesigameko ngasinye.
I-OpenTelemetry njengolimi oluvamile lokubona
Enye yezindlela ezicacile ukuhlanganiswa kwe I-OpenTelemetry (OTel) njengendlela ejwayelekile ye-telemetry evulekileLuhlaka lomthombo ovulekile oluchaza ama-API, ama-SDK, kanye nezingxenye ukuze kuqoqwe amamethrikhi, amalogi, kanye nokulandelelwa ngendlela efanayo, ngaphandle kokuxhunyaniswa nomkhiqizi wamathuluzi athile okubonwa.
Eminyakeni ezayo, kulindeleke ukuthi Izinkampani zifuna ukuhambisana ne-OpenTelemetry kubathengisi bayo. Isizathu silula: ngokusebenzisa "ulimi olujwayelekile" ukuchaza i-telemetry, inhlangano ingashintsha amapulatifomu okubona ngaphandle kokubhala kabusha noma ukusebenzisa kabusha yonke ikhodi yayo. Lokhu kunciphisa ingozi yokukhiya kwabathengisi futhi kunikeza ukuguquguquka kokuguqula isitaki njengoba kudingeka.
Ngokungafani nezixazululo eziphelele, lapho ukuhlanganiswa ngakunye okusha kuncike kumephu yomenzi, i-OTel Ivumela ukuhlanganiswa ukuthi kusinde izinguquko zobuchwepheshe.Njengoba kuvela izinsizakalo ezintsha zamafu, amafreyimu, noma izikhathi zokusebenza, zidinga nje ukukhipha i-telemetry ngefomethi ejwayelekile ukuze zikwazi ukuyithumela kunoma iyiphi i-backend ehambisanayo.
Ngaphezu kwalokho, ukusetshenziswa kwe-OpenTelemetry kubalulekile ekuqinisekiseni ukuthi ukondla kahle ubuhlakani bokwenziwaAmamodeli e-AI, kungaba ukufunda komshini kwendabuko, ukutholwa kwe-anomaly, noma i-AI yokukhiqiza, asebenza kahle kakhulu uma idatha ihlanzekile, ihlelekile, futhi ihambisana. I-OTel inikeza ngqo uhlaka olufanayo lokukhiqiza nokulebula i-telemetry ezocutshungulwa ama-algorithms.
Izifundo zakamuva ziphakamisa ukuthi izinhlangano esezisebenzisa i-OpenTelemetry kakadeNgisho noma zisetshenziswa kancane, zibona umthelela omuhle ezinkomba ezifana nokukhula kwemali engenayo, izinzuzo zokusebenza ezithuthukisiwe, kanye nedumela lomkhiqizo. Akuyona imilingo: ukuba nesisekelo sokubonakala esihambisanayo nesiphathekayo kwenza kube lula ukubona izinkinga ngaphambi kokuba zithinte ikhasimende futhi kuthuthukiswe ukusebenza kwezinsizakalo ezibalulekile.
Izinsika ezintathu zomkhuba wanamuhla wokubuka
Ngaphandle kokwamukela indinganiso efana ne-OTel, umkhuba wokubona okuzwakalayo uncike ku izingxenye ezintathu eziyisisekelo eziqinisana: izinsimbi ezivulekile, izinto ezixhunyiwe (noma idatha), kanye nokukwazi ukuhlela.
La izinsimbi ezivulekile Lokhu kuhilela ukuqoqa i-telemetry kusuka kuma-ejenti azimele kanye nalawo avulekile. Izinhlelo zokusebenza, izinsizakalo, ababungazi, izitsha, imisebenzi engenaseva, izinhlelo zokusebenza zeselula, izinsizakalo zamafu eziphethwe—konke kumele kukwazi ukukhipha amamethrikhi, imicimbi, amalogi, kanye nokulandelelwa ngamafomethi angalinganiswa. Yilapho ama-ejenti avela kubathengisi bendabuko eqala khona ukusebenza, kodwa futhi nabathumeli kanye nemitapo yolwazi evela ku-OpenTelemetry nakwamanye amaphrojekthi avulekile.
Ibhlogo lesibili yilelo le- izinhlangano ezixhunyiwe kanye nemethadathaUkuqoqa nje amamethrikhi namalogi akwanele; udinga ukuqonda ukuthi ubani owakhiqizayo nokuthi ahlobene kanjani. Lokhu kudinga ukuhlonza izinsizakalo, izizindalwazi, imigqa, imisebenzi, ama-pod, amaqoqo, ama-akhawunti amafu, nokuxhumanisa i-telemetry kanye nokuncika kwawo. Ngalesi simo, ipulatifomu ingenza ngokuzenzakalelayo amamephu okwakha, ukugeleza kwezingcingo, kanye nezikhathi zezehlakalo ngaphandle kokuthi ithimba lilungiselele konke ngesandla.
Ngokusekelwe kulokho, umuntu angafaka isicelo ubuhlakani kanye nokuhlaziya okuthuthukisiweNgokubona amaphethini, ukungalingani, kanye nokuxhumana ngaphakathi kwesethi yedatha, amapulatifomu okubona angasiza ekubekeni phambili izexwayiso, ukunciphisa umsindo, ukuthola izehlakalo eziyinkimbinkimbi, nokusheshisa ukuhlaziywa kwembangela yezimbangela. Lena yindlela yemvelo eya ekubonweni okusebenzayo futhi, njengoba sizobona kamuva, eya ekuzimeleni kwe-ejenti.
Ekugcineni kukhona ukuhlelekaIbhizinisi ngalinye linezidingo ezithile: ama-KPI alo, izinqubo ezibucayi ezahlukene, kanye namamodeli ezindleko ahlukile. Ipulatifomu yesimanje yokubona kufanele ivumele ukwakha izinhlelo zokusebenza ezenziwe ngokwezifiso kanye nemibono ngaphezu kwayo yonke i-telemetry: amadeshibhodi ahlanganisa idatha yobuchwepheshe nezilinganiso zebhizinisi, ukuhlaziywa komthelela wezomnotho kokuphazamiseka noma ukuwohloka, noma izinhlelo zokusebenza zangaphakathi zokuphenya izehlakalo eziyinkimbinkimbi ngokuya ngomsebenzi wenkampani.
Leli khono "lokuhlela" idatha yokubona livula umnyango wokusebenzisa izimo ezifana nokuthi linganisa izindleko zangempela zephutha Enkambisweni yokukhokha, yihlobanise nembangela yobuchwepheshe (isibonelo, ukuhlehla kwesevisi encane yokukhokha) bese ubeka phambili imizamo yokulungisa ngezinqubo zomthelela kwezomnotho kuphela.
Ukubonwa okugxile ebhizinisini: kusukela kukhonsoli kuya emphumeleni
Enye yezinguquko ezinkulu ezilindelwe ukushintsha kusuka kolunye ukubonwa okugxile ekusebenzeni kobuchwepheshe kwenye egxile ebhizinisini ngokucacile. Idatha efanayo—amalogi, imikhondo, izilinganiso, izenzakalo—iqala ukusetshenziswa hhayi nje kuphela ekulondolozeni ingqalasizinda, kodwa futhi nasekusetshenzisweni phendula imibuzo ebalulekile mayelana nemali engenayo, izindleko, kanye nolwazi lomsebenzisi.
Emikhakheni yezimboni, isibonelo, ukubonwa kwezinzwa ze-IoT kuvumela bikezela ukwehluleka kwemishini futhi uthuthukise izinhlelo zokulungisa. Uma kutholakala amaphethini okudlidliza angajwayelekile noma amazinga okushisa angaphandle kwebanga, ukungenelela kungahlelwa ngaphambi kokuba umugqa wokukhiqiza ume, ukuvimbela isikhathi sokungasebenzi esingahleliwe kanye nemiphumela yako yezomnotho.
Emkhakheni wezezimali, ukuhlaziya ngesikhathi sangempela amalogi okuthengiselana Kuyasiza ekuboneni ukuthengiselana okusolisayo okungase kuhlobane nokukhwabanisa. Uma uhlelo luthola ukulandelana kwemicimbi okungajwayelekile, izindawo ezingavamile, noma amanani ahlukana namaphethini ajwayelekile, kungabangela izindlela zokuvimba ezizenzakalelayo noma ukubuyekezwa ngesandla ngaphambi kokuba ukuhlasela kuphumelele.
Ekukhangiseni nasekuthengiseni, ukuhlanganisa ukulandelwa kwezinhlelo zokusebenza ngezilinganiso zomkhankaso Ikuvumela ukuthi uphendule imibuzo eqondile kakhulu: Ingabe ukubambezeleka kwewebhusayithi kuthinta izinga lokuchofoza noma ukuguqulwa? Yiluphi uhlobo lwesici oluthuthukisa kangcono ukuzulazula nesikhathi sokuhlala? Uma ukusebenza kwehla ngesikhathi somkhankaso, ukubonwa kusiza ekuboneni ukuthi zingaki izintengiso ezingaba khona ezilahlekile nokuthi inkinga yenzeke kuliphi iphuzu eliqondile ku-funnel.
Konke lokhu kuhilela ukuhumusha i-telemetry yobuchwepheshe ibe yi- ulwazi olusebenzayo lwabaholi bebhizinisiAkukhona ukubonisa umqondisi wokuthengisa igrafu ye-CPU, kodwa ukubabonisa ukuthi zingaki izinkokhelo ezihlulekile ukuqedwa ngenxa yokuwohloka kwesevisi nokuthi izindleko ezilinganisiwe beziyini. Futhi ukuze kufezwe lokhu, ukubonwa kufanele kuxhume idatha yobuchwepheshe, imicimbi yomsebenzisi, kanye nezilinganiso zebhizinisi ngaphakathi kwemodeli efanayo.
Izinkampani zokubonisana ezigxile ekuqapheleni, njengeNettaro, sezivele zisiza izinkampani nezikhungo ukuthi ukwenza lokhu kusuka embonweni osebenzayo kuphela kuya embonweni osuhlelwe kahleukuklama amamodeli axhumanisa ama-KPI ebhizinisi nezimpawu ze-telemetry zesikhathi sangempela.
Kusukela ku-AIOps kuya ku-Agent Observability
Ukwamukelwa kwe Ubuhlakani Bokwenziwa kumapulatifomu okubonwa Sekuvele kuyiqiniso. Amaqembu amaningi e-ITOps afake izingxenye ze-AIOps—ama-algorithms ahlaziya inani elikhulu ledatha yokusebenza ukuthola okungahambi kahle, imicimbi yamaqembu, noma ukubikezela izinkinga—emisebenzini yawo yokusebenza.
Ezimweni eziningi, kuyahlanganiswa futhi I-AI ekhiqizayo ukusebenzisana ne-telemetry usebenzisa ulimi lwemvelo: buza imibuzo yengxoxo efana nokuthi "kungani amaphutha angu-500 anda eYurophu emizuzwini engu-20 edlule?" bese uthola incazelo esekelwe kumarekhodi, izilinganiso, kanye nemikhondo ngaphandle kokwakha imibuzo eyinkimbinkimbi.
Kodwa-ke, izinqumo eziningi namuhla zisekelwe ku-AI Ziyaqhubeka nokubuyekezwa ngabantuAma-algorithm asiza ukuhlunga umsindo nokuhlonza izimbangela ezingaba khona, kodwa amaqembu okusebenza agcina ukulawula, aqinisekisa izincomo, futhi enze ngesandla izenzo eziningi zokulungisa. Ukuzethemba okuphelele ezinqumweni ezenzakalelayo kusalokhu kulinganiselwe.
Yilapho-ke Ukuqashelwa Kwe-ejentiLena indlela lapho ama-ejenti e-AI ethatha indima yokuzimela kakhulu: awagcini nje ngokuthola amaphethini nokuchaza okwenzekayo, kodwa futhi Baphatha imisebenzi ephelele, kusukela ekuboneni iphutha kuya ekusebenziseni ikhambi elifanele.
Kulo modeli, i-ejenti ingakwazi, isibonelo, ukubona ukwanda okungavamile kokubambezeleka kwesevisi ebalulekile, ikuhlobanise nokusetshenziswa okuthile, ihlole umlando wezigameko ezifanayo, bese izinqumela ukuthi ngabe qala i-rollback, sikala umthamo, noma sebenzisa ukucushwa okuhlukileKonke lokhu kubhalwe phansi ngokuningiliziwe ukuze kuhlolwe futhi kubuyekezwe abantu okungenzeka kamuva.
Njengamanje, yizinkampani ezimbalwa kuphela ezisebenzisa lokhu Ukubonwa Kwe-Agent Esebenzayongokulungiswa okuzenzakalelayo kanye nokubikezela izinkinga okuthuthukile. Kodwa izibikezelo zibonisa ukuthi ukwamukelwa kwayo kuzokhula kakhulu, kuqhutshwa ukufuna umkhiqizo omkhulu emaqenjini e-IT kanye nesidingo sokunciphisa isikhathi abasichitha emisebenzini yokulungisa ephindaphindwayo.
Imikhawulo yokuqondisa ngesandla kanye nesidingo sokuzimela
Isidingo sama-ejenti azisebenzelayo siqondwa kangcono uma sibheka izimo ezimbi kakhulu njenge- ukubonwa kwemodeli yolimi olukhulu (i-LLM)Ukuqapha ngesandla lezi zinhlobo zezinhlelo kuwumsebenzi ocishe ungenzeki: amanani edatha makhulu kakhulu, izakhiwo zihlanganisa izingxenye eziningi ezisatshalaliswe, futhi isidingo sokuqapha ngesikhathi sangempela sihlala njalo.
Ubuningi bamarekhodi kanye nezilinganiso kwenza kube Ukuhlonza izinkinga ngesandla kuhamba kancane kakhuluNoma yikuphi ukubambezeleka ekutholeni ushintsho ekuziphatheni, ukwanda kwamaphutha, noma ukwehla kwekhwalithi yezimpendulo kungaba nemiphumela emibi ezindaweni zokukhiqiza, kokubili maqondana nolwazi lomsebenzisi kanye nedumela kanye nokuhambisana nemithetho.
Ngaphezu kwalokho, ukubheka ngesandla kudla izinsiza eziningi zabantu; ukuthambekela emaphutheni futhi akukhuli kahle Njengoba inani lamamodeli, izimo, noma ukuhlanganiswa nezinhlelo zokusebenza zebhizinisi likhula, okungase kusebenze kuhlelo lokuhlola olunabasebenzisi abambalwa kuba yisithiyo lapho uhlelo luqaliswa kuyo yonke inhlangano.
Ngakho-ke, ezindaweni eziyinkimbinkimbi njengalezo ezihilela i-LLM noma izakhiwo ezisabalele kakhulu, isidingo izixazululo zokubuka ezizimeleSikhuluma ngezinhlelo ezikwazi ukuhlaziya njalo i-telemetry, ukubona ukuphambuka, ukuphakamisa noma ukwenza izinyathelo zokulungisa, nokufunda ekungeneleleni ngakunye ukuze kuthuthukiswe ukusebenza kwazo ngokuhamba kwesikhathi.
Ama-ejenti esenzo sokubona kanye nokwenza ngokuzenzakalela ku-interface
Ukuthuthuka kwe-AI akugcini nje endaweni yokubonakala "kwakudala". Ucwaningo olwenziwe yizinkampani ezifana ne-NVIDIA, ngamaphrojekthi afana ne- I-NitroGen Kuyimodeli eqhubayo ehlanganisa amakhono okubona kanye nesenzo: ama-ejenti abuka isikrini, anqume isimo sendawo futhi anqume ukuthi yini okufanele ayenze ngokulandelayo, ngaphandle kokuhlanganiswa okuqondile nesistimu ayilawulayo.
Ngobuchwepheshe, lokhu kuhilela ukuqeqesha imodeli nge iqembu elikhulu lamavidiyo emidlalo noma ukusebenzisana ukuze bafunde ukuhlobanisa lokho abakubonayo nezenzo uchwepheshe angazenza. Basebenza ngokulandelana kwesikhathi, ukuhlukanisa ukunyakaza, imigomo yesikhathi eside, kanye nokwenza ngcono ngaphansi kwemingcele eminingi efana nokubambezeleka noma ukuzinza.
Nakuba isibonelo esibonakala kakhulu imidlalo, le ndlela yombono-isenzo inamandla amakhulu ebhizinisini: ivumela ukudalwa ama-ejenti asebenza ku-interface yesithombe okuvamile, ukuzulazula izinhlelo zokusebenza eziyinkimbinkimbi, ukusebenzisa ukugeleza okuphindaphindiwe, ukuqinisekisa izinqubo, noma ukwenza ukuhlolwa kokuphela ngaphandle kwesidingo sama-API athile.
Lokhu kumelela uhlobo lokuvela kwemvelo kwe-RPA yendabuko eya ku- Ukuzenzakalelayo okuhlakaniphile, okunesimo esingokomqondoAmacala avamile okusetshenziswa afaka ukuhlolwa kwesofthiwe okuzenzakalelayo okulingisa ukuziphatha kwangempela komsebenzisi, ukwesekwa okuqondiswayo okuphindaphinda lokho okufanele kwenziwe yisisebenzi ngokuchofoza ngakunye, ukukhiqizwa kwedatha yokwenziwa ye-QA, noma "amawele edijithali" aphinda umsebenzi womuntu ezinhlelweni zenkampani.
Ukuze konke lokhu kube nokwenzeka, uhlaka oluqinile lokuphepha kwe-inthanethi, ukuphatha, kanye nokubukaAma-ejenti asebenzisana nezixhumi ezibalulekile kanye nezinhlelo kumele alandele izinqubomgomo zokufinyelela, agweme izenzo eziyingozi, abhale phansi zonke izinyathelo ngezinjongo zokuhlola, futhi asebenze ngaphakathi kwemingcele echazwe ngokucacile. Ukubonwa lapha kusebenza njengebhokisi elimnyama kanye "nebhokisi lamathuluzi": kurekhoda lokho okwenziwa yi-ejenti futhi kunikeze idatha yokulinganisa nokuthuthukisa ukuziphatha kwayo.
Ezokuphepha, ukubusa, kanye ne-Zero Trust enkathini yama-ejenti e-AI
Ukwanda kwe-AI ejenti kanye nezinhlelo ezizimele kuletha lokhu Izingozi ezintsha okumele ziphathwe ngokucophelelaEnye yezinto okuxoxwe kakhulu ngazo yi-"shadow AI": ama-ejenti, amamodeli noma ukuhlanganiswa okuqaliswa ngaphandle kweziteshi ezisemthethweni zenhlangano, ngaphandle kokuphepha okwanele noma ukulawulwa kokuthobela imithetho.
Kukhona futhi ingozi yokuthi ama-ejenti amabili noma ama-ejenti anonyaLokhu kungenzeka ngokuklama (ukuhlasela kwangaphandle, ukushintshwa okusheshayo, ukufakwa kwemiyalelo) noma ngenxa yamaphutha okucushwa avumela uhlelo olunezinhloso ezinhle ukuthi lwenze izenzo ezingahlosiwe. Ukuze kuncishiswe lezi zingozi, kubalulekile ukusebenzisa izimiso ze I-Zero Trust ikakhulukazi maqondana nobuhlakani bokwenziwa.
I-Zero Trust kulo mongo isho ukuthi Akukho agent noma ingxenye ye-AI ebhekwa "njengothembekile" ngokuzenzakalelayo.Yonke into kumele ivunyelwe ngokusobala, izimvume kumele zikhawulelwe kokuncane okudingekayo (isimiso selungelo elincane), futhi konke ukuxhumana kumele kubhalwe phansi ukuze kuhlolwe kamuva. Ngakho-ke ukubonwa kuba yinto ebalulekile ekubusweni kwe-AI.
Ukuba nokuqaphela okuhle kuvumela ukuqapha ngesikhathi sangempela lokho okwenziwa ama-ejenti, ukutholakala kokuziphatha okungajwayelekile, ukuqinisekiswa kwezinqubomgomo zokufinyelela, kanye nokutholakala kobufakazi obuphelele uma kwenzeka izehlakalo. Amathuluzi afana nohlu lwezenzo ezivunyelwe, ukubuyekezwa kwabantu kwezihibe ezibucayi, ukuhlanzwa kwedatha ebucayi, kanye nokulawula indawo yokubala (esakhiweni, ifu lomphakathi, ifu elizimele) kuyizinto ezibalulekile zohlu lokuhlola oluqinile. ukuphathwa kwe-AI okusebenzayo.
Kulesi simo, kubalulekile ukuthola ibhalansi phakathi kokusungula izinto ezintsha nokulawulaIzinhlangano zifuna ukusebenzisa ngokugcwele amandla e-AI e-ejensi ukuze zithole umkhiqizo kanye nokuncintisana, kodwa ngaphandle kokudela ukuphepha, ukuthobela imithetho, noma ukucaca ekwenzeni izinqumo okuzenzakalelayo.
Idatha, ingqalasizinda, kanye ne-AI njengesendlalelo esiyisisekelo sebhizinisi
Uma sibheka isithombe esikhulu, i-AI iyashintsha isuka ekubeni ithuluzi elengeziwe iye ekubeni yithuluzi elibalulekile. ungqimba lwesakhiwo lapho ukuncintisana kwezomnotho kusekelwe khonaKonke kugxile kulolo shintsho: amasu edatha, ukwakheka kwamafu, ukwakheka kwehadiwe, amamodeli abasebenzi, ngisho nezinqubomgomo zikazwelonke mayelana nengqalasizinda yedijithali.
Ngakolunye uhlangothi, Idatha ihlanganiswa njengesihlukanisi esiyinhloko sokuncintisanaNjengoba ukubala kanye nokumodela kuba yinto ethengiswa kakhulu, okwenza umehluko ukuba nedatha yakho esezingeni eliphezulu, ephethwe kahle. Ukuqaphela, ngokubamba i-telemetry ecebile neyomongo, kuba ngomunye wemithombo ewusizo kakhulu yedatha izinhlelo ze-AI zamandla futhi kuthuthukiswe izinqubo.
Ngakolunye uhlangothi, i- Ingqalasizinda ye-AI isiqala ukubonwa njengempahla yezwe ehlelekileUkwanda kwamafu azimele kusabela esidingweni sokulawula lapho idatha ebucayi igcinwa futhi icutshungulwa khona, ukuthi amamodeli aqeqeshwa kanjani, nokuthi asebenza ngaphansi kwaziphi izinhlaka zomthetho. Amazwe atshala imali ezikhungweni zedatha ezilungiselelwe imisebenzi ye-AI, ezisebenzisa amandla kahle, futhi ezihambisana nezidingo zokuthobela imithetho.
Konke lokhu kuhambisana ne- ukusheshisa ukwenziwa kwesimanje kwezikhungo zedathaIcindezelwe yizidingo zamandla nokupholisa zemithwalo yemisebenzi ye-AI kanye nezinhlelo zama-ejenti, ukusebenza kahle kwamandla akuseyona nje inkinga yokusebenza kodwa sekuyinto evimbelayo yokusungula izinto ezintsha kanye nesidingo sokuthobela imvelo.
Ngesikhathi esifanayo, izinkampani ziphoqelekile ukuthi ukuqeqesha kabusha abasebenzi bayoUmgomo akukhona ukuguqula wonke umuntu abe ngumhleli wezinhlelo, kodwa ukuqeqesha ochwepheshe abakwazi ukuhlela nokusebenzisa lezi zinhlelo ezizimele: ochwepheshe bebhizinisi abasebenzisa i-AI, onjiniyela abangahumusha izidingo zokusebenza zibe yizinqubomgomo zokubuka kanye nezokuphepha, kanye nezindima ezixubile eziqonda kokubili umthelela wezobuchwepheshe kanye nowezomnotho wezinqumo.
Uma kubhekwa konke, lokhu kuthuthuka kuholela esimweni lapho ukubonwa okuvulekile nokuzimela kakhudlwana Kuba yinto ehlanganisa ubuchwepheshe, ibhizinisi kanye nomthethonqubo: amazinga afana ne-OpenTelemetry aqinisekisa ukuthuthwa kwedatha kanye nekhwalithi, i-AI kanye ne-Agent Observability kunciphisa ubunzima bokusebenza futhi kusheshiswe impendulo yezehlakalo, kanti ukubusa kanye nemikhuba ye-Zero Trust kuqinisekisa ukuthi konke lokhu kwenzeka ngaphansi kokulawulwa, ngokuphephile nangokuhlonza kwangempela.
Izinhlangano ezikwazi ukuchaza le nhlanganisela - i-telemetry ejwayelekile, amapulatifomu ahlanganisiwe, ukugxila emiphumeleni yebhizinisi, kanye nama-ejenti e-AI alawulwa ngokubona okuhle - zizoba sesimweni esihle sokuncintisana endaweni lapho izinhlelo zedijithali ziqhubeka nokuba bucayi, ziyinkimbinkimbi, futhi zizimele, kodwa futhi zikwazi ukukhiqiza inani elibonakalayo uma zilawulwa ngokubonakala okufanele.
