- Ukubonwa kwe-AI kwandisa izingodo zakudala, izilinganiso kanye nemikhondo enezimpawu ezithile ze-AI njengokukhukhuleka, ubuthi, ukubona izinto ezingekho kanye nomthelela webhizinisi.
- Imodeli enezigaba ihlanganisa i-telemetry, ukuhlolwa kwekhwalithi, umjikelezo wokuphila kanye nokuphatha, kanye nokuphepha kanye nezindleko njengezinto ezithinta yonke indawo.
- Abashayeli bezindiza abangochwepheshe be-AI kanye ne-GenAI badinga ukulandelela okujulile, kwe-ejenti ngayinye kanye nokwenza ngokuzenzakalelayo okuhlakaniphile ukuze kugcinwe ubunzima bulawuleka.
- Amapulatifomu ahlangene, imikhuba ye-SRE kanye nezilinganiso ze-AI ezithembekile kubalulekile ekukhuliseni i-AI ngokuphephile kuwo wonke amafu, ukuphepha kanye nokusebenza kwebhizinisi.
Izinhlelo ze-AI zeqe umugqa kusukela kuma-prototype okuhlola kuya kwingqalasizinda ebalulekile yebhizinisi, futhi lokho kushintsha imithetho yomdlalo yokuqapha nokulawula. Uma amamodeli amakhulu olimi (ama-LLM), imisebenzi ye-ejenti noma ama-copilot akhiqizayo ethinta uhambo lwamakhasimende, imali engenayo noma ukuphepha, opharetha abasakwazi ukuthembela ku-Application Performance Monitoring (APM) yendabuko kuphela. Badinga isu lokubona elihlukaniswe ngezigaba eliveza ukuthi lezi zinhlelo ezingaba khona, ezivame ukungacaci, ukuthi kungani ziziphatha ngaleyo ndlela nokuthi zithinta kanjani yonke ingxenye ye-stack.
Lesi sihloko sigxila kakhulu ezingqimbeni ezibalulekile zokuqaphela i-AI, sihlanganisa imibono evela ekuqapheleni kwamafu, i-SRE, imisebenzi yezokuphepha kanye ne-AI ephethe kahle ibe umbono owodwa novumelanayo. Sizohamba ngesisekelo se-telemetry, ukuhlolwa kwekhwalithi okuqhubekayo, ukuphathwa kokuguquguquka kanye nomjikelezo wokuphila, ukubusa kanye nokulandelela, kanye nezidingo ezikhethekile zabashayeli bezindiza be-AI kanye ne-GenAI abangochwepheshe. Endleleni, uzobona ukuthi ukuqaphela kokubili ngoba AI kanye nge I-AI ishintsha imisebenzi, kusukela ezinkampanini ezintsha zaseLatin America ezikhulisa ama-LLM kuya emabhizinisini omhlaba wonke athola amafu ahlanganisiwe.
Kusukela ku-APM yakudala kuya ekubonweni kwe-AI egcwele
Sekungamashumi eminyaka amaqembu okusebenza ethembele kumathuluzi e-APM ukugcina ama-monolith kanye nezinhlelo zokusebenza ezisatshalaliswa ekuqaleni ziphilile, kodwa izakhiwo zesimanje ezisebenzisa i-AI ziye zadlula leyo modeli. Ezindaweni zendabuko, ikhodi isetshenziswa emijikelezweni ebikezelwayo, ukuncika kuyaqondakala kahle futhi ama-KPI afana nokuphuma, izinga lamaphutha kanye nokusetshenziswa kwe-CPU kuvame ukwanele ukuthola nokulungisa izinkinga zokusebenza.
Ukuguqulwa kwedijithali kanye namaphethini endabuko efwini kuye kwandisa kakhulu ubunzima ngisho nangaphambi kokuba i-AI ingene esithombeni. Ama-microservices kuma-Kubernetes clusters, imisebenzi engenaseva ephila ama-millisecond kanye nezinsizakalo ze-polyglot ezikhipha ama-log ngezindlela ezahlukene konke kukhiqiza amavolumu amakhulu e-telemetry okungasakwazi ukuwabamba ngokunembile amasampula e-minute level. Ukuqaphela kwavela ukuze kungene ama-metrics aphezulu, imicimbi, ama-log kanye nama-trace (MELT) esikalini futhi kuhlobanise ngesikhathi sangempela.
Manje engeza ama-LLM, isizukulwane esitholiwe esingeziwe (i-RAG) kanye nama-ejenti azimele ngaphezu kwaleyo ndwangu esivele iyinkimbinkimbi, futhi inselele yokubonakala iba bukhali nakakhulu. Lezi zinhlelo zethula ukunganqumi, ukuziphatha okuphuthumayo, ukuhamba komsebenzi okuqhutshwa ngokushesha kanye nokushintshashintsha kwemodeli, okungekho nokukodwa okubonakala ngokucacile kugrafu elula ye-HTTP latency. Udinga ukubonwa okuqonda amathokheni, izixwayiso, izihlungi zokuphepha, izindleko ngombuzo ngamunye kanye nomthelela wezinga lebhizinisi.
Ngamafuphi, ukubonwa kwe-AI akuyona indawo ehlukile, kodwa ukunwetshwa kokubonwa kwesimanje okwengeza izimpawu ezithile ze-AI phezu kwedatha ye-MELT ekhona. Inhloso isafana—ukuphendula okuthi “Kwenzekani, kungani, futhi yini okufanele siyenze?”—kodwa imibuzo kumele ibuzwe kuwo wonke amamodeli, ama-ejenti, amapayipi edatha, ingqalasizinda kanye nemiphumela yabasebenzisi ngesikhathi esisodwa.
Isigaba 1: Izilinganiso ze-Telemetry kanye nengqalasizinda
Isisekelo sanoma yiliphi isu lokubona izinto ezingabonakali yi-telemetry eqinile: ama-metric, ama-log kanye ne-traces echaza indlela i-AI stack yakho eziphatha ngayo ngesikhathi sokusebenza. Ngemithwalo yemisebenzi ye-AI, lokho kusho ukudlulela ngale kwamashadi e-CPU ajwayelekile kanye nememori nokuqoqa amasignali aqaphela imodeli ahlobene ngqo nokusebenza kanye nezindleko.
Ezingeni lengqalasizinda, usadinga amamethrikhi akudala njengokulibaziseka, ukudlula kanye nokusetshenziswa kwezinsiza, kodwa kufanele uwalandelele ngobuningi bezingxenye ze-AI. Lokho kufaka phakathi ukusetshenziswa kwe-GPU ngemodeli ngayinye, ingcindezi yememori yedathabheyisi ye-vector, amazinga esicelo namaphutha ama-endpoints okuphetha kanye nezinkomba zokugcwala kwezinqubomgomo zokukala ngokuzenzakalela ku-AWS, i-Azure noma amanye amafu. Ukuhlanganisa ukwanda kwethrafikhi nezilinganiso zengqalasizinda yamafu kubalulekile lapho umsebenzi we-AI ukhula ngokushelela.
Kuma-LLM ikakhulukazi, i-telemetry yezinga le-token iba isakhamuzi sesigaba sokuqala. Abaqhubi kufanele baqophe amathokheni asheshayo, amathokheni okuqeda kanye namathokheni aphelele ngocingo ngalunye, kanye nesikhathi sokuphendula, inguqulo yemodeli kanye nohlelo lokusebenza lokushaya ucingo. Ngenxa yokuthi ama-LLM amaningi ezentengiselwano akhokhiswa ngethokheni ngalinye, lokhu kuqondanisa kuyisisekelo sokuqonda nokulawula izindleko ngombuzo ngamunye, izindleko ngesici ngasinye kanye nezindleko ngesigaba ngasinye sekhasimende.
Ukulandelela okusatshalaliswayo nakho kudinga ukwandiswa ukuze kuhlangatshezwane nezingcingo ze-AI, hhayi nje kuphela ama-web endpoints kanye nemibuzo yedathabheyisi. Ukulandelela kufanele kufake phakathi ama-span esicelo ngasinye se-LLM, isicelo samathuluzi, isinyathelo sokubuyisa noma ucingo lwangaphandle lwe-API olusetshenziswa yimodeli. Ngaleyo ndlela, lapho i-latency igxuma, amaqembu angabona ukuthi inkinga itholakala yini ekuthokheni, ekubhekeni okushumekiwe, ku-node ye-GPU egcwele ngokweqile noma ku-API yenkampani yangaphandle ehamba kancane.
Ukuhlanganisa le telemetry ecebile nge-AI namapulatifomu okuqapha amafu akhona kuletha i-AI engxoxweni efanayo yokusebenza njengayo yonke ingxenye ye-stack. Uma ukukhishwa okusha kubangela amazinga aphezulu amaphutha esangweni le-API kanye nokwanda kokusetshenziswa kwethokheni le-LLM, ukubonwa okuhlanganisiwe kukhombisa ukuthi lezi yizinhlangothi ezimbili zesigameko esifanayo kunokuba kube yiziphambeko ezihlukile.
Isigaba 2: Ukuhlolwa okuqhubekayo kwekhwalithi yokukhipha i-AI
Uma i-telemetry eyisisekelo isikhona, ungqimba olulandelayo lugxila kulokho okuhlukanisa ngempela ukubonwa kwe-AI nokuqapha okuvamile: ukuhlolwa okuqhubekayo kwekhwalithi yokukhipha imodeli. Izinhlelo ze-AI zingase zisheshe futhi zishibhile kodwa zibe yingozi uma ziphupha, zivuza idatha noma zingaqondi kahle inhloso yomsebenzisi.
Izilinganiso zekhwalithi ze-AI kumele zichazwe ngamagama agxile ebhizinisini esikhundleni samaphuzu okunemba kobuchwepheshe kuphela. Kumsizi wentengiselwano, lokho kungaba ukunemba kwezinguquko ze-oda noma ukubuyiselwa kwemali; komshayeli wesibili wokusekela, izinga lesisombululo kanye nokwaneliseka; kwenjini yokuncoma, ukufaneleka kanye nokuchofoza. Lawa ma-KPI ahumusha okulindelwe yisizinda kube yizimpawu ezibonakalayo.
Ngenxa yokuthi imiphumela ye-LLM iwulimi lwemvelo, ukuhlolwa kwekhwalithi kuvame ukuhlanganisa ukwahlulela komuntu nezilinganiso ezisizwa yi-AI. Amaqembu angagcina amasethi edatha egolide—izimpendulo ezibhalwe ngochwepheshe ezikhuthazweni ezingokoqobo—futhi aqhathanise njalo izimpendulo zemodeli ezibukhoma nalezo zinkomba. Ngesikhathi esifanayo, angasebenzisa amagreyidi asekelwe kumodeli ukuthola amaphuzu ngesisekelo, ukufaneleka, ukuhambisana, ukushelela kanye nokunamathela kumongo womthombo.
Izilinganiso zengozi nokuphepha zifanelwe ukugqanyiswa kwazo ohlakeni lokuhlola. Amapayipi okubonwa kufanele alandelele ukuthi izihlungi zokuqukethwe zivimba kangaki izikhuthazo noma ukuqedwa ngenxa yobudlova, ukuzilimaza, inkulumo enenzondo noma izihloko ezibucayi, nokuthi yiziphi izimo zokusetshenziswa ezibangela lezi zinkinga kakhulu. Ukwanda kokuqukethwe okuvinjiwe kungabonisa imizamo yokufaka ngokushesha, ukushintsha kwesizinda noma ukuvimbela okunganele.
Amasu asekelwe kuma-ejenti kanye nokulingisa asiza ekukhuliseni ukuhlolwa ngale kwemiyalelo elula. Ngokuzenzakalela izingxoxo ezishintshashintshayo phakathi kwama-ejenti noma phakathi komsebenzisi wokwenziwa kanye nohlelo lwe-AI, amaqembu angahlola amacala onqenqema, izimo zokubuyela emuva kanye nokuziphatha komongo omude ngaphambi kokuba kuthinte abasebenzisi bokukhiqiza. Lokhu kunamandla kakhulu emisebenzini eyinkimbinkimbi ye-ejenti, lapho isinqumo esisodwa esibi ekuqaleni kochungechunge singasakazeka ngezingcingo eziningi zamathuluzi.
Isigaba 3: Ukutholwa kokukhukhuleka kanye nokuphathwa komjikelezo wokuphila we-AI
Ngisho nemodeli eziphatha kahle ngosuku lokuqala ingaba engathembekile ngokuhamba kwesikhathi uma idatha, ukuziphatha komsebenzisi noma uhlelo oluzungezile kushintsha—yilapho ukutholwa kokuzulazula kanye nokuphathwa komjikelezo wokuphila kungena khona. Ngaphandle kokubonwa okucacile kokushelela, amaqembu avame ukuqaphela sekwephuzile ukuthi ukusebenza kuye kwehla, ngemva kokuba abasebenzisi sebevele bezwe umthelela.
Ukuqapha ukuzulazula kwedatha kuqala ngokulandelela izakhiwo zezibalo zokufakwayo ngokuhamba kwesikhathi bese uziqhathanisa nokusabalalisa okusetshenziswa ngesikhathi sokuqeqeshwa kanye nokuqinisekiswa kokuqala. Ukushintsha kolimi, amakhathalogi omkhiqizo, imigomo yokulawula noma izibalo zabantu kungabangela amamodeli ukuthi aqonde kabi imibuzo noma abuyele ezimpendulweni ezijwayelekile nezingenalusizo. I-Telemetry kufanele ibambe izici ezifana nokuvama kwesizinda, ukusatshalaliswa kwezinhlangano noma amaphethini ajwayelekile okukhuthaza.
Ukuzulazula kwemodeli kudlula okungenayo futhi kubheka izinguquko emiphumeleni noma ezinqumweni, noma ngabe idatha engenayo ifana. Ukuqaphela kufanele kulinganise ukunemba, ukucwasa, ubuthi kanye nezinye izindlela zokulinganisa ikhwalithi ngezigaba, kugqamise lapho ukuziphatha kwemodeli kuhluke khona kunesisekelo sayo. Lokho kungabonakala njengokubona izinto ezingekho endaweni ethile, noma amazinga okwenqaba akhuphukayo kumaphrofayili athile amakhasimende.
Ama-loop empendulo avela kubasebenzisi bokugcina ayisibonakaliso esibalulekile kulolu ngqimba. Izilinganiso ezilula zesithupha phezulu/phansi, impendulo yombhalo okhululekile kanye nokuhlelwa komsebenzisi kwemidwebo eyenziwe nge-AI konke kwembula ukuthi uhlelo lusaletha yini inani. Amapulatifomu okubuka kufanele aphathe lezi zimpawu njengezilinganiso zekilasi lokuqala futhi azifake ekuqeqeshweni kabusha noma ekulungiseni kahle amapayipi.
Ukuze kusebenze impendulo yokukhukhuleka, izexwayiso kumele zixhumane ngqo nemisebenzi yokusebenza yomjikelezo wokuphila njengokuqeqeshwa kabusha, ukukhushulwa kwemodeli noma ukubuyiselwa emuva. Uma ukuzulazula kudlula imingcele evunyelwene ngayo—ake sithi, ukulahlekelwa okunembile okungaphezu kuka-5-10% uma kuqhathaniswa nesisekelo—izindlela zokulinganisa zingabangela ukuqoqwa kwedatha, ukuhlolwa okusha kuqhubeke, futhi, ngemva kokuqinisekiswa kuphela, ukukhishwa kwamamodeli abuyekeziwe. Lokhu kuvala umjikelezo phakathi kokutholwa nokulungiswa ngaphandle kokuthembela kuphela kumaqhawe enziwe ngesandla.
Isigaba 4: Ukulandelelwa, ukubusa kanye ne-AI ethembekile
Njengoba izinhlelo ze-AI zihlangana nomthethonqubo, ubumfihlo kanye nokuziphatha, ukubonwa kumele futhi kuhlinzeke ngamakhono aqinile okulandelela kanye nokuphatha. Akusanele ukwazi ukuthi “imodeli ishilo”; izinhlangano zidinga ukuchaza ukuthi yiziphi izimpendulo, izincomo, amamodeli kanye nokuhlelwa okuholele emiphumeleni ethile.
Ukubhalwa kokufakwayo kanye nokuphumayo kusukela ekuqaleni kuze kube sekupheleni, kanye nezinguqulo zamamodeli kanye namathempulethi asheshayo, kuyinsika yokulandelelwa kwe-AI. Yonke indlela yokwenza izinqumo—kusukela embuzweni womsebenzisi kuya ekutholeni, ekwakheni ngokushesha, ekubizeni amathuluzi kanye nempendulo yokugcina—kufanele ikwazi ukwakhiwa kabusha kusukela kumarekhodi. Lokhu kubalulekile ekuhlolweni kwamabhuku, ekuphenyweni kwezigameko kanye nasekuphenduleni imibuzo yokulawula mayelana nokwenza izinqumo okuzenzakalelayo.
Ukuphatha akukhona nje ukuqopha imibhalo; kumayelana nokuphoqelela izinqubomgomo zokufinyelela, ukugcinwa kanye nokusetshenziswa kwedatha ebucayi. Izitolo zokubuka kumele zihlangane nokuphathwa kobunikazi nokufinyelela, ukubethela kanye nokufihla idatha, ukuqinisekisa ukuthi imisebenzi egunyaziwe kuphela engahlola amalogi athile noma idlale ukusebenzisana okubucayi. Lokhu kubaluleke kakhulu emikhakheni engaphansi kwe-GDPR, i-HIPAA noma imithetho yezezimali.
Izimiso ze-AI ezithembekile—ubulungiswa, ukucaca, ukuzibophezela, ubumfihlo, ukuphepha kanye nokubandakanya—zidinga ama-proxies abonakalayo ohlelweni. Amamethrikhi alandelela okuqukethwe okulimazayo, ukuchezuka kwabantu, ukuphika okungachazeki noma ukuvimba ngokweqile ngezihlungi kunikeza indlela yokulinganisa yokusebenzisa lezi zimiso ngokoqobo. Izexwayiso ezihambisana nalezi zinkomba zingakhuthaza ukubuyekezwa kwabantu ngaphambi kokuba umonakalo odalwe yidumela noma wezomthetho uqongeleleke.
Kubathengisi besofthiwe abazimele (ama-ISV) abakha ama-copilot noma izici ze-GenAI zamakhasimende, ukubonwa nakho kusekela izivumelwano zezinga lesevisi abanganikeza ngokuthembekile. Ama-SLO mayelana nokubambezeleka, ukutholakala, amazinga ezigameko zokuphepha kanye nama-KPI ebhizinisi athembele ku-telemetry ethembekile kanye nekhono lokufakazela ukuthobela imithetho ngokuhamba kwesikhathi.
I-Agent AI: Ukuqashelwa kwemisebenzi yama-ejenti amaningi
Imboni ishintsha ngokushesha isuka ezimweni zokusetshenziswa kwe-LLM ezisheshayo iye kwi-AI e-agent, lapho ama-ejenti amaningi ehlangana khona, ebiza amathuluzi futhi ehlanganisa ndawonye—intuthuko yamakhono ehambisana nokukhula kobunzima. Ukulungisa noma ukulawula lezi zinhlelo ngamalogi ajwayelekile cishe akunakwenzeka; aziphathi njengama-API aqondile kodwa zisebenza njengemisebenzi eguquguqukayo, esakazwayo.
Kuhlelo lokusebenza olujwayelekile lwe-ejenti, isicelo ngasinye somsebenzisi singase sibangele izendlalelo eziningana zomsebenzi: i-orchestration logic, ukuncenga ama-ejenti amaningi, izingcingo zamathuluzi, ukuzama kabusha, ukwenza ngcono kanye namagatsha okuphatha amaphutha. Ngaphandle kokubonakala okucacile, amaqembu abona kuphela isicelo se-HTTP sangaphandle, aphuthelwe nhlobo ukuthi iyiphi i-ejenti eyenze siphi isinqumo, ngokulandelana kwanoma yimuphi umongo.
Ukulandelela izinga le-ejenti kugcwalisa lesi sikhala ngokwabela ama-span hhayi nje kuphela kumasevisi, kodwa nakuwo wonke ama-ejenti kanye nokushaya kwamathuluzi. Abaqhubi bathola imephu yokusebenzisana kwama-ejenti amaningi: ukuthi yimaphi ama-ejenti ayehilelekile, ukuthi adlulise kanjani umongo, ukuthi asebenza kuphi ngesikhathi esifanayo nokuthi kwavela kuphi izithiyo noma ukwehluleka. Leyo mephu iba ithuluzi eliyinhloko lokuhlaziya imbangela yezimpande lapho izincomo zihamba kancane noma zingalungile.
Izindaba zangempela zibonisa ukuthi lokhu kubaluleke kangakanani. Cabanga ngethimba lobunjiniyela be-e-commerce elakha injini yokuncoma eqhutshwa yi-AI enama-ejenti akhethekile: elilodwa lokusesha umkhiqizo, elinye lokuhlaziya imizwa ekubuyekezweni kanye nelesithathu lokwenza okunikezwayo kube ngokwakho. Lapho izincomo ziqala ukubuyisa imiphumela engabalulekile noma ebambezelekile, ngaphandle kwemininingwane eqashelwa yi-ejenti, ukulungisa amaphutha kuba umsebenzi wokuqagela. Ngokubona okugcwele kwe-AI, ithimba lingabona, isibonelo, ukuthi i-ejenti yokwenza kube ngokwakho ilinde ngokuphindaphindiwe i-API yephrofayela yangaphandle ehamba kancane, noma ukuthi i-ejenti yokwenza kube ngokwakho iphelelwa yisikhathi emibhalweni emide yokubuyekeza.
Amapulatifomu asekela ngokujwayelekile ukubonwa kwe-ejensi—ama-ejensi okumaka, amathuluzi kanye nobudlelwano bawo—avumela amaqembu ukuthi asuke ekucimeni umlilo aye ekuthuthukisweni okuhlelekile. Ziqokomisa amathuluzi angasetshenziswa kahle, izinto ezinomsindo, izindawo zokwehluleka njalo kanye namathuba okwenza ngcono ukufana noma ukugcinwa kwemininingwane. Lokhu kuwukubonwa okuklanyelwe ngokuqondile i-AI, hhayi ukuvuselelwa kusukela ekulandeleni okuvamile.
I-AI yokubona izinto kalula: imisebenzi ehlakaniphile, yokuxoxa
Olunye uhlangothi lwemali lusebenzisa i-AI uqobo ukuguqula indlela amaqembu asebenzisa ngayo idatha yokubona, eshintsha esuka kumadeshibhodi asabelayo aye emisebenzini yokuxoxa esebenzayo. Izitaki zanamuhla zikhiqiza i-telemetry eningi kunanoma yimuphi umuntu angayihlola ngokunengqondo; ama-LLM kanye nama-ejenti angasiza ekuqondeni ngesikhathi sangempela.
Izixhumi kanye nezinqubo ze-ejenti yokungaqondisi abathengisi zenza kube nokwenzeka ukuveza idatha yokubona ngqo kunoma yini onjiniyela abasizi be-AI abayisebenzisayo kakade. Esikhundleni sokuphoqa amaqembu ukuthi ashintshe izimo phakathi kwama-IDE, ama-chatbot kanye nama-UI okuqapha, i-ejenti yokubona ingadalula amamethrikhi namalogi nge-interface ejwayelekile i-GitHub Copilot, i-ChatGPT, i-Claude noma amanye amathuluzi angayibuza.
Empeleni, lokhu kusho ukuthi onjiniyela bangabuza imibuzo yolimi lwemvelo njengokuthi “Bekuyini izinga lethu lamaphutha kusukela ekusetshenzisweni kokugcina?” noma “Ngiboniseni okungahambi kahle ku-LLM latency ehoreni elidlule” bese bethola izimpendulo eziqhutshwa idatha ngaphandle kokushiya indawo yabo yokusebenza eyinhloko. Izexwayiso, izifinyezo zezehlakalo kanye nemibiko yezitayela konke kungakhiqizwa futhi kulungiswe ngengxoxo, kunciphisa isithiyo sokungena kwamalungu eqembu angachwepheshile kangako.
Izinhlangano ezifaka ukubonwa kubasizi bazo be-AI zibika ukuthi kunesikhathi esisheshayo sokuxazulula inkinga (MTTR) kanye nokukhathala okuncane kokushintsha umongo. Uma ithimba lobunjiniyela lezinkundla zokuxhumana, isibonelo, lingabuza ngempilo yokukhiqiza ngaphakathi komsizi ofanayo abalisebenzisayo ukubhala nokubukeza ikhodi, impendulo yesigameko iba ukugeleza okukodwa okuqhubekayo esikhundleni sokusebenza okuqhekekile kwethuluzi.
Uma kuqhathaniswa nezindlela ezidinga ukucushwa okunzima ngesandla, njengamaphakheji amakhono akhiwe ngesandla, ukuhlanganiswa okuguquguqukayo, okusekelwe kuphrothokholi kunciphisa ukungqubuzana futhi kuvumela amaqembu ukuthi asebenzise amathuluzi amaningi e-AI ngesikhathi esisodwa. Lokhu kugcina onjiniyela belawula ukukhetha kwabo amathuluzi ngenkathi besahlanganisa idatha yokubona, okuyibhalansi ebalulekile yezinhlangano eziqaphela ukuvalelwa kumthengisi oyedwa we-AI.
Ukuqaphela ukuphepha: ukubona izinsongo ngesikhathi sangempela
Amaqembu ezokuphepha abhekene noshintsho olufanayo: ukuqapha okuvamile kanye nezixazululo ze-SIEM ziyazabalaza ukuhambisana nomthamo, ubuhlakani kanye nesivinini sezinsongo zanamuhla, ikakhulukazi ezindaweni eziqhutshwa yi-cloud-first, i-AI. Ukuqaphela ukuphepha kwandisa ingqondo yokuqaphela ingozi kanye nokusabela ezigamekweni, okunikeza ukuqonda okujulile nokuqhubekayo ngalokho okwenzekayo kuwo wonke ama-endpoints, amanethiwekhi, ubunikazi kanye nezinhlelo zokusebenza.
Ngokungafani nokuqapha okusekelwe emkhawulweni okuphakamisa ama-alamu kuphela lapho izimo ezichazwe kusengaphambili zephulwa, ukuqaphela ukuphepha kuhlose ukwakha kabusha izindlela zokuhlasela eziyinkimbinkimbi kusukela ku-telemetry eningiliziwe. Ihlobanisa amasignali avela kuma-endpoints, amaseva, izinsizakalo zamafu kanye nokuziphatha komsebenzisi ukuze kutholakale ukungalingani okucashile—ukunyakaza okuseceleni, ukusetshenziswa kwamalungelo okungavamile, ukufinyelela idatha okusolisayo—okungabonakali kuma-log afihliwe.
Isikhathi sokuxazulula siyisilinganiso esibalulekile lapha: izinhlangano eziningi zibika amanani ajwayelekile e-MTTR angaphezu kwehora ngezinkinga zokukhiqiza, okuyinto engamukeleki kakhulu uma kubhekwa izindleko zesikhathi sokungasebenzi kanye nokulahleka kwedatha. I-Telemetry ethembekile kakhulu, ukuhlaziywa okuhlanganisiwe kanye nokuxhumana okuzenzakalelayo kusiza ukunciphisa lelo thuba, okuvumela amaqembu ukuthi asuke ophenyweni lwangemva kokufa aye ekugcinweni kwezindiza.
Izingxenye eziyinhloko zokubonwa kwezokuphepha zibonisa ukubonwa okuvamile kodwa zinoshintsho olugxile engcupheni. Ukuqoqwa kwe-Telemetry kuhlanganisa izindawo zokugcina, ukugeleza kwenethiwekhi, izindiza zokulawula amafu kanye nabahlinzeki bobunikazi; ukuhlanganiswa kwamalogi kulungisa amafomethi ahlukahlukene; ukulandelela kwakha kabusha izindlela zesicelo; ukuhlaziya okuthuthukisiwe kanye nokubuka kokufunda komshini kwamaphethini abonisa ukuhlaselwa; kanye namadeshibhodi aphakathi aveza ukuma kokuphepha okuphelele, kwesikhathi sangempela.
Amapulatifomu esimanje e-SIEM ne-XDR athuthukisiwe nge-AI ahlanganisa le ndlela, ahlanganisa idatha ehlelekile nengahlelekile ibe ngamachibi edatha angalinganiswa futhi abeke phezulu imisebenzi yokutholwa okuzenzakalelayo, uphenyo kanye nokuphendula. I-Hyperautomation ithatha indawo yezincwadi zokudlala ze-SOAR eziphukile, ezithungwe ngesandla, kuyilapho isavumela ukubusa kwabantu ngaphezu kwezenzo ezithinta kakhulu. Lokhu kuhlanganiswa kuthuthukisa ukunemba kokutholwa, kunciphisa umsindo futhi kusiza amaqembu okuphepha ukugxila ezenzakalweni ezibaluleke kakhulu.
Imikhuba emihle kakhulu yokufeza ukubonwa kwe-AI kusukela ekuqaleni kuze kube sekupheleni
Ukwakha ukubonwa kwe-AI okuphelele kumayelana nenqubo kanye namasiko njengoba kunjalo nangamathuluzi, futhi imikhuba embalwa esebenzayo ivele njalo ekusetshenzisweni okuphumelelayo. Ukuphatha ukubonwa njengento edingekayo yesigaba sokuqala kusukela esigabeni sokuklama, kunokuba kube yinto ecatshangelwe kamuva, kuyindlela eyodwa ebaluleke kakhulu yokushintsha indlela yokucabanga.
Okokuqala, chaza amamodeli e-telemetry acacile ahlanganisa ingqalasizinda, ukuziphatha okusebenzayo kanye nomthelela webhizinisi. Ngasohlangothini lwengqalasizinda, nquma ukuthi ungalinganisa kanjani ukubambezeleka, ukudlula kanye nokusetshenziswa kwezinsizakusebenza zengxenye ngayinye ye-AI. Ngasohlangothini lokusebenza, khetha izilinganiso ezifana nokunemba, amazinga okubona izinto ezingekho, izinkomba zokubandlulula noma izihlungi zokuphepha. Ngasohlangothini lwebhizinisi, landelela ukuguqulwa komsebenzisi, isikhathi esilondoloziwe, izindleko ngokusebenzisana ngakunye noma ukufinyelelwa kwe-SLA.
Okwesibili, hlanganisa ukumuncwa kwedatha kanye nokuxhumana kwayo ukuze zonke izimpawu ezihlobene ne-AI—ezobuchwepheshe, ezokuphepha, ibhizinisi—zikwazi ukuhlaziywa ndawonye. Ukuletha izibalo, izingodo, imikhondo kanye nemicimbi yokuphepha echibini elilodwa lokubona kuvumela imibuzo ehlukahlukene efana nokuthi “Ingabe lesi senzakalo sokukhukhuleka sihambisane nokuphazamiseka kokuphepha?” noma “Lowo modeli omusha uthinte kanjani kokubili izindleko kanye nezikhathi zokuxazulula izinkinga?”
Okwesithathu, yenza ngokuzenzakalelayo konke okuphephile ngangokunokwenzeka: ukuxwayisa, ukubona izinto ezingavamile, ukucebisa izehlakalo, kanye nezimpendulo lapho kufaneleka khona. Ukuhlaziywa okusekelwe ku-AI kungagqamisa izinto ezingaphandle kwemithombo ye-metric, kufinyeze izehlakalo, kuphakanyiswe izinyathelo zokulungisa futhi kuqhutshwe nezenzo ezinobungozi obuphansi ngokuzenzakalelayo. Abaphenduli abangabantu bese begxila ekuhluleleni, ekuhwebelaneni okuyinkimbinkimbi kanye nasekuthuthukisweni kwesikhathi eside.
Okwesine, tshala imali emakhono eqembu kanye nokuqondana okwabiwe. Ukuqaphela kusebenza kakhulu uma onjiniyela, ososayensi bedatha, ama-SRE, abahlaziyi bezokuphepha kanye nabanikazi bemikhiqizo bonke bazi ukuthi bangahumusha kanjani amadeshibhodi, izexwayiso kanye nokulandelwa. Ukuqeqeshwa, imibhalo kanye nokubuyekezwa kwezigameko ezisebenza ngezindlela ezahlukene kusiza ekwakheni ulimi oluvamile mayelana nempilo ye-AI kanye nengozi.
Okokugcina, qaphela izindleko kanye nobumfihlo ngenkathi wandisa ukumbozwa kokubuka. I-Telemetry ayikhululekile, futhi ukuqoqwa kwedatha okunamandla kungadala izinselele zokuthobela imithetho. Ukusampula okuhlakaniphile, izinqubomgomo zokugcina amazinga kanye nokulawula ukufinyelela okuqinile kuqinisekisa ukuthi ukubonwa kuhlala kuzinzile futhi kuhambisana nezibopho zomthetho.
Ukuhlanganisa lezi zingqimba ndawonye—ubuchwepheshe be-telemetry, ikhwalithi, ukuzulazula, ukubusa, ukulandelela ama-ejenti, ukuphepha kanye nemisebenzi esizwa yi-AI—kuguqula i-AI isuke ebhokisini elimnyama elingacacile, elibuthakathaka ibe yingxenye ehlolwayo, elungisekayo yebhizinisi lakho ledijithali, okwenza amaqembu akwazi ukuhamba ngokushesha ngokuzethemba kunokuba abe nethemba.
