Wani sabon aikin buɗe tushen yana son ɗaukar nazari zuwa matakin gaba kuma shine mutanen da ke baya BlazingSQL kwanan nan ta sanar da cewa ta saki lambar tushe don injin SQL, wanda ake amfani dashi a GPU don saurin sarrafa bayanai. BlazingSQL ba cikakken DBMS bane, amma an sanya shi azaman injin don bincika da aiwatar da manyan bayanan bayanai, kwatankwacin ayyukanta zuwa Apache Spark.
Ga wadanda basu san BlazingSQL ba su sani wannan injin GPU ne wanda aka inganta shi akan tsarin halittu na RAPIDS wanda saiti ne na dakunan karatun laburare na software don gudanar da bincike na karshen-zuwa-karshen da bututun kimiyyar bayanai akan GPUs.
A cewar kungiyar, An kirkiro BlazingSQL ne don magance tsada, sarkakiya da saurin tafiyar da masu amfani ke fuskanta yayin aiki a manyan majalisu na bayanai. BlazingSQL ya dace don yin tambayoyin bincike na mutum akan manyan bayanan bayanai (dubun gigabytes) waɗanda aka adana a cikin tsarin layi (misali rajistan ayyukan, ƙididdigar NetFlow, da sauransu).
Don aiki tare da GPU, ana amfani da saiti na dakunan karatu na RAP abWasu sun haɓaka tare da sa hannu na NVIDIA, suna ba ku damar ƙirƙirar sarrafa bayanai da aikace-aikacen bincike waɗanda ke gudana gaba ɗaya a kan hanyar GPU (ana ba da hanyar Python don amfani da ƙananan matakan CUDA da ƙididdigar layi ɗaya).
BlazingSQL yana ba da ikon amfani da SQL maimakon API cuUDF sarrafa bayanai (dangane da Kibiya Apache) wanda RAPIDS ke amfani dashi. BlazingSQL ƙarin layin ne wanda ke gudana a saman cuDF kuma yana amfani da laburaren cuIO don karanta bayanai daga faifai.
Tambayoyin SQL sune fassara zuwa kiran aikin cuUDF, wanda ke ba da damar shigar da bayanai akan GPU kuma yi aiki tare, tarawa, da kuma sarrafa ayyukan akan su. Yana goyan bayan ƙirƙirar abubuwan rarrabuwa waɗanda suka shafi dubban GPUs.
Amfani da SQL yana ba RAPIDS damar haɗawa tare da tsarin nazari na yanzu ba tare da masu sarrafa rubutu ba ba tare da yin amfani da shigar da bayanai zuwa matsakaicin DBMS ba, yayin ci gaba da dacewa tare da dukkan bangarorin RAPIDS, fassara ayyukan da ake ciki a cikin SQL da kuma tabbatar da aikin kwod. Ya haɗa da tallafi don haɗakawa tare da ɗakunan karatu na XGBoost da cuML don warware bincike da ayyukan koyon injin.
BlazingSQL iya gudanar da tambayoyi daga fayilolin lebur a cikin sifofin CSV da Apache Parquet wanda ke cikin hanyar sadarwa da tsarin girgije kamar HDSF da AWS S3, kai tsaye suna canja sakamakon zuwa ƙwaƙwalwar GPU.
Godiya ga ayyukan daidaitawa a kan GPU da kuma amfani da saurin ƙwaƙwalwar bidiyo, aiwatar da tambaya a cikin BlazingSQL ya ninka sau 20 fiye da na Apache Spark.
BlazingSQL yana sauƙaƙa sauƙaƙa aiki tare da bayanai - maimakon ɗaruruwan kiraye-kirayen aikin cuDF, zaka iya yin sa ta hanyar tambaya guda SQL.
"BlazingSQL yana magance wadannan damuwar kwastomomin ba kawai tare da saurin wuce gona da iri ba, injin SQL GPU da aka rarraba, har ma da himma kan sauki," Rodrigo Aramburu, Shugaba na BlazingSQL, ya rubuta a cikin wani shafin yanar gizo na gaba. "Tare da 'yan layuka kaɗan, BlazingSQL na iya tambayar ɗanyen bayanan ku, duk inda yake zaune, kuma yana iya yin hulɗa tare da RAPIDS ɗin ku da kuma bayanan nazari."
BlazingSQL yana bawa masu amfani damar tambaya data data lake data kafa kai tsaye a cikin ƙwaƙwalwar GPU azaman GPU DataFrame (GDF). GDF aiki ne wanda ke ba da goyan baya don aiki tare tsakanin aikace-aikacen GPU. Hakanan yana ma'anar takaddun bayanan GPU na gama gari.
"Ta hanyar amfani da Aprow Arrow akan GPUs da haɗawa tare da Dask, BlazingSQL zai faɗaɗa aikin buɗe tushen kuma ya tura madaidaicin aiki tare cikin saurin yanayin kimiyyar bayanai."
Ga masu sha’awa su sani an rubuta lambar a cikin C ++ tare da haɗin kera don masu amfani kuma tushen buɗewa yana ƙarƙashin lasisin Apache 2.0.