Hologres Uyasebenza

I-Hologres iyinkundla ehlanganisiwe ye-Agentic yokusesha nokuhlaziya idatha enezinhlobo eziningi eyethulwe yi-Aliyun Cloud. Ihambisana nephrothokholi ye-PostgreSQL, isekela ukubhalwa nokubuyekezwa kwedatha ngesikhathi sangempela ngezinga le-PB, ukuhlaziya kwe-OLAP/okwesikhashana, kanye nokusesha kwamaphoyinti anokuncintisana okuphezulu. Inkundla isebenzisa i-SQL ehlanganisiwe ukuhlanganisa idatha ehlelekile, ehleleke kancane kanye nedatha engahlelekile ye-OSS. Ihlanganisa i-Object Table, i-Dynamic Table ne-AI Function ukuze inikeze amandla okusesha umbhalo ogcwele, okusekelwe kuma-vector, okuxubile kanye nokubiza amamodeli, iphinde inwebe iye kwimemori yesikhathi eside yama-Agent kanye nesisekelo sokubonakala kokusebenza, okuyenza ifanele izimo ze-Data+AI/RAG.

I-OLAP yesitolo sedatha sesikhathi sangempela: Isekela ukubhalwa kwedatha ngezinga le-PB ngesikhathi sangempela, ukubuyekezwa, ukuhlaziya okunezinhlangothi eziningi kanye nemibuzo yesikhashana.
I-Serving yokusesha kwamaphoyinti anokuncintisana okuphezulu: Ihambisana nephrothokholi ye-PostgreSQL, futhi yakhelwe imibuzo eku-inthanethi enokubambezeleka okuphansi kanye nezinsizakalo zedatha ezinokuncintisana okuphezulu.
Ukusesha umbhalo ogcwele/ama-vector/okuxubile: Isethi eyodwa yedatha isekela ukusesha umbhalo ogcwele, ukusesha okusekelwe kuma-vector, ukusesha okuxubile kanye nokubuyisa kwe-RAG.
Ukucubungula kwe-Data+AI enezinhlobo eziningi: Sebenzisa i-Object Table ukuxhuma amafayela angahlelekile e-OSS, i-Dynamic Table ekucubunguleni okwandayo, kanye ne-AI Function ekubizeni amamodeli ukuze kuhlaziywe, kusikwe izingxenye, kwenziwe i-Embedding, kuhunyushwe futhi kukhiqizwe okuqukethwe.
Ukuhlanganiswa kwe-lakehouse nemithombo eminingi: Ingaxhuma ku-MaxCompute, Paimon, Iceberg, Flink, DataWorks kanye nemithombo yedatha yangaphandle efana ne-MySQL/PostgreSQL.
Umsizi we-AI wesitolo sedatha: Uhlanganisa ulwazi lokuxilonga, ukukhiqiza i-SQL, ukuncoma ukucaciswa kanye nokuphathwa kwezindleko lube ama-Skills asebenzisekayo.

Vula isilingo: Vakashela iwebhusayithi esemthethweni ye-Aliyun Hologres ku-https://cn.aliyun.com/product/hologres, chofoza ukuvula/ukuzama umkhiqizo bese uqedela ukuqinisekiswa kwe-akhawunti nokukhokhisa.
Dala i-instance: Dala i-instance ye-Hologres, ukhethe isifunda, uhlobo lwenethiwekhi, ukucaciswa kwe-CU kanye nendlela yokugcina.
Lungisa ukufinyelela: Lungisa i-VPC/inethiwekhi yomphakathi, uhlu olumhlophe lwamakheli e-IP, i-akhawunti yedathabheyisi kanye nezimvume ezincane ezidingekayo.
Xhuma ku-instance: Sebenzisa i-HoloWeb, i-psql, i-JDBC/ODBC noma ithuluzi le-BI ukuxhuma ku-instance bese udala idathabheyisi.
Yakha imodeli namathebula: Dala amathebula ngokomthwalo womsebenzi, uhlukanise isitoreji semigqa nesamakholomu, bese wengeza lapho kudingeka khona ukhiye oyinhloko, ukhiye wokusabalalisa, inkomba yama-vector kanye nenkomba yombhalo ogcwele.
Faka idatha: Sebenzisa i-INSERT/COPY, i-Flink, i-DataWorks, i-MaxCompute noma ithuluzi lokuvumelanisa ukufaka idatha ku-Hologres.
Phatha idatha enezinhlobo eziningi: Dala i-Object Table ehambisana namafayela e-OSS, sebenzisa i-Dynamic Table ekucubunguleni okwandayo phezu kwe-lake, bese usebenzisa i-AI Function ekuhlaziyeni, ekusikeni izingxenye, ku-Embedding noma ekukhiqizeni.
Umbuzo nokuhlaziya okuhlanganisiwe: Sebenzisa i-SQL ehlanganisiwe ukwenza ukusesha kwamaphoyinti, i-OLAP, ukusesha umbhalo ogcwele/okusekelwe kuma-vector/okuxubile noma inference ye-AI, bese ugcina ama-view avame ukusetshenziswa nemibuzo egcinwe ngokuqinile.
Hlola umthwalo bese wethula: Ngaphambi kokwethula, hlola ukuncintisana, ukubambezeleka, ukunemba kokubuyisa, umthamo wokungenisa kanye nezindleko zama-Token zemodeli.
Ukusebenza nokwenza ngcono: Ngemva kokuqalisa ekukhiqizeni, vula ukuqapha kwemibuzo ehamba kancane, ukusetshenziswa kwezinsiza, izindleko kanye nokuhlolwa, bese ulungisa izinsiza ngokuguquguqukayo lapho kudingeka.

Isakhiwo esihlanganisiwe: Ihlanganisa i-OLAP, ukusesha kwamaphoyinti, umbhalo ogcwele, ama-vector, ukusesha okuxubile kanye ne-inference ye-AI enjinini eyodwa, kunciphise ukuhlanganiswa kwezingxenye eziningi ezifana ne-ClickHouse/Greenplum + HBase/Redis + Elasticsearch.
Idatha eyodwa isetshenziswa ezimweni eziningi: Isethi efanayo yedatha isebenzela imibiko ye-BI, ukuhlaziya kwesikhashana, izinsizakalo eziku-inthanethi kanye nokusesha kwe-RAG/Agent, kunciphise ubungozi bamaketanga okuvumelanisa nokungahambisani kwedatha.
Ukusebenza kwesikhathi sangempela okuqinile: Isekela ukubhalwa ngesikhathi sangempela, ukubuyekezwa ngesikhathi sangempela kanye nokusesha ngokushesha ngemva kokubhala, ifanele izimo ezihlanganisa i-Serving enokuncintisana okuphezulu nokubambezeleka okuphansi kanye nokuhlaziya kwesikhathi sangempela.
I-Data+AI eyakhelwe ngaphakathi: Nge-Object Table, i-Dynamic Table ne-AI Function, amafayela e-OSS, ukucubungula okwandayo phezu kwe-lake kanye nokubiza amamodeli kufakwa ochungechungeni lwe-SQL.
Ukuhlanganiswa okuhle ne-lakehouse kanye ne-ecosystem ye-Aliyun Cloud: Ixhumana kahle ne-MaxCompute, Flink, DataWorks, OSS, Bailian nezinye izingxenye, futhi ifanele ukuqaliswa ngaphakathi kohlelo lwe-Aliyun Cloud.
Ukuphathwa okuphelele nokuguquguquka: Ukuphathwa kwe-cloud-native kwehlisa izindleko zokusebenza, futhi izinsiza zingalungiswa ngokuvumelana nomthwalo, okuyenza ifanele amabhizinisi anokuguquguquka okukhulu kwevolumu yedatha nokuncintisana.

I-BI yesikhathi sangempela nokuhlaziya ukusebenza kwebhizinisi: Bhala idatha yokwenziwe, yama-oda kanye neyethrafikhi ngesikhathi sangempela ukuze usekele amadeshibhodi emibiko, ukuhlaziya kwesikhashana kanye nokuqapha kwebhizinisi endaweni eyodwa.
Ukulawula ubungozi nokumaketha ngesikhathi sangempela: Hlanganisa izehlakalo zesikhathi sangempela, amaphrofayela abasebenzisi kanye nokubuyisa okusekelwe kuma-vector/umbhalo ogcwele ukuze wenze izinqumo zobungozi ezingama-millisecond, useshe amacala afanayo futhi wenze izincomo ezenzelwe umuntu ngamunye.
Isizindalwazi solwazi se-RAG yebhizinisi: Ngemva kokuhlaziya ama-PDF/izithombe/amavidiyo anomsindo, sika izingxenye bese wenza i-Embedding, usebenzise ukusesha okuxubile ekuphenduleni imibuzo, ekusizeni amakhasimende nasekusesheni izinqubomgomo.
Ukuhlaziya izingcingo zamakhasimende namathikithi omsebenzi: Hlukanisa izikhulumi ekurekhodweni, wenze i-ASR transcription, ukuhlaziya imizwa nokukhiqiza isifinyezo, bese ugcina imiphumela njengempahla yedatha eseshwayo.
Imemori yesikhathi eside yama-Agent nokuhlolwa kwezinhlobo eziningi: Nikeza izimo ezifana nokufunda okusizwa yi-AI, i-AI Coding kanye nokuhlolwa kwamashalofu amandla ememori ewela izikhathi zokuxhumana, ukuqhathanisa izithombe kanye nokuqinisekisa umehluko.