Iyo nyowani yakavhurwa sosi chirongwa inoda kutora analytics kune inotevera nhanho uye ndeyekuti vanhu vari kumashure BlazingSQL ichangobva kuzivisa kuti yakaburitsa kodhi yekodhi yeyayo SQL injini, iyo inoshandiswa muGPUs kuti ikurumidze kugadzirisa data. BlazingSQL haisi DBMS yakazara, asi yakamisikidzwa seinjini kuongorora uye kugadzirisa makuru data maseti, akafanana mumabasa ayo kuna Apache Spark.
Kune avo vasingazive BlazingSQL vanofanira kuziva izvo iyi injini yekumhanyisa SQL yakavakirwa pane RAPIDS ecosystem inova seti yeakavhurika sosi software raibhurari yekumhanyisa magumo-kumagumo analytics uye data sainzi mapaipi pane maGPU.
Zvinoenderana nechikwata, BlazingSQL yakagadzirirwa kugadzirisa mari, kuomarara, uye kunonoka kunosangana nevashandisi pavanoshanda mumisangano mikuru ye data. BlazingSQL yakakodzera kuitisa yega mibvunzo yekuongorora pane makuru data maseti (makumi egigabytes) akachengetwa mumatafura mafomati (semuenzaniso matanda, nhamba dzeNetFlow, nezvimwewo).
Kushanda neGPU, seti yemaraibhurari eRAPIDS anoshandiswa abZvimwe zvakagadzirwa nekubatanidzwa kweNVIDIA, zvichikubvumidza iwe kuti ugadzire kugadzirisa data uye kuongorora zvinoshandiswa zvinomhanya chose kudivi reGPU (chinongedzo chePython chinopihwa kushandisa yakaderera-chikamu CUDA yekutanga uye kuverenga kwakaenzana).
BlazingSQL inopa kugona kushandisa SQL pane iyo API cuUDF dhata yekugadzirisa (yakavakirwa paApache Museve) inoshandiswa neRAPIDS. BlazingSQL ndeimwe yekuwedzera iyo inomhanya pamusoro pecuDF uye inoshandisa iyo cuIO raibhurari kuverenga data kubva ku disk.
SQL mibvunzo iri translate mu cuUDF basa mafoni, ayo anotendera data kuti iiswe kune iyo GPU uye ita merge, aggregate, uye sefa mashandiro pavari. Inotsigira kugadzirwa kwekuparadzaniswa kwekugadziriswa kunotenderera zviuru zveGPUs.
Kushandiswa kwe SQL inobvumira RAPIDS kuti ibatanidzwe nemaitiro epo aripo asina kunyora ma processor usingashandisi yepakati kurodha pasi dhata mune imwezve DBMS, uku uchichengetedza kuenderana kuzere nezvikamu zvese zveRAPIDS, kushandurira mashandiro aripo muSQL uye kuona cuDF-nhanho yekuita. Inosanganisira rutsigiro rwekubatanidzwa neXGBoost uye cuML maraibhurari ekugadzirisa ongororo uye mabasa emuchina ekudzidza.
BlazingSQL inogona kumhanya nemibvunzo kubva kumafaira akapfava mu CSV uye Apache Parquet mafomati iri pane network uye makore masisitimu akadai seHDSF neAWS S3, ichinyatso kuendesa mhedzisiro kune memory yeGPU.
Kutenda kune yekufananidza mashandiro pane iyo GPU uye nekushandisa nekukurumidza vhidhiyo ndangariro, kubvunza kuurayiwa muBlazingSQL inokwana kusvika makumi maviri nekukurumidza kupfuura muApache Spark.
BlazingSQL inonyanya kurerutsa kushanda nedata - pachinzvimbo chemazana eiyo cuDF basa rekufona, unogona kuzviita uine imwechete SQL mubvunzo
"BlazingSQL inotaura nezvekunetsekana kwevatengi kwete neinokurumidza kusimba, yakaparadzirwa SQL GPU injini, asi zvakare kutarisisa kushingairira," Rodrigo Aramburu, CEO weBlazingSQL, akanyora mune inotevera blog. "Nemitsetse mishoma yekodhi, BlazingSQL inogona kubvunza data rako mbishi, kwese kwarinogara, uye nekudyidzana neRAPIDS yako iriko uye siteki ye analytics."
BlazingSQL inogonesa vashandisi kubvunza bhizinesi dhata data data rinoisa zvakananga muGPU memory seGPU DataFrame (GDF). GDF chirongwa chinopa rutsigiro rwekubatana pakati peGPU zvinoshandiswa. Iyo zvakare inotsanangudza yakajairika GPU memory data layer.
"Nekushandisa Apache Museve pane maGPU uye kusanganisa neDask, BlazingSQL ichawedzera kuvhurwa sosi mashandiro uye kutyaira inotevera wave yekushandirana mune inomhanyisa data data sainzi ecosystem."
Kune avo vanofarira vanofanira kuziva izvozvo kodhi yacho yakanyorwa muC ++ ine python interface yevashandisi uye yakavhurwa sosi iri pasi peApache 2.0 rezinesi.