MapD: Database wanda ke aiki akan GPUs

A yau mun sami kanmu muna fuskantar abin mamaki na Big Data, zamu iya samun adadi mai yawa daga maɓuɓɓuka marasa iyaka. Wannan adadi mai yawa yana kawo fa'idodi da yawa, amma kuma yana kawo ƙalubale da yawa. Mafi na kowa daga cikinsu: lokutan amsawa a cikin babban kundin bayanai.

TaswiraD an haife shi don bayar da saurin gudu a fagen bayanan bayanan nazari. Tsara don aiki tiriliyoyi na rikodin a cikin batun milliseconds amfani da ikon sarrafa kwamfuta da aka bayar ta GPUs. An gina shi daidai don amfani da duk kayan aiki da damar software da ke cikin katunan zane, yana ba manazarta da masanan kimiyyar bayanai lokutan martani game da umarni uku na girma (x3) sama da fasahohin da aka taɓa amfani da su don waɗannan dalilai. Amfani da kwatankwacin GPUs (Kusan kusan 1000 a GPUs na zamani) da manyan bandwidth na ƙwaƙwalwar ajiya (Around 80000Gbps) don aiwatar da ayyukan aljebra na layi da bincike na bayanai, ta amfani da LLVM don tarawa a ainihin lokacin kowace tambaya, ban da adana bayanan da aka nema sosai. a cikin ɓoye na GPUs (ƙwaƙwalwar DDR8 mai saurin gaske).

Dole ne mu tuna cewa a cikin duniyar Big Data, ba a amfani da rumbunan bayanan gargajiya, gwargwadon rubutu da adana fayiloli, tunda waɗannan za su haifar da yawan ayyuka na I / O a kan rumbun kwamfutarka. Don manufar nazarin biliyoyin bayanai, da a-ƙwaƙwalwar ajiya, kamar Apache Spark. Koyaya, don samun adadin ƙwaƙwalwar ajiyar da ake buƙata da aikin da ake buƙata, ƙungiyar uwar garken ya zama dole kuma mun san cewa wannan yana haifar da tsada a cikin kayan aiki, kebul na cibiyar sadarwa da mafi yawan masu fasaha. Saboda haka, TaswiraD yana ba da damar cimma babban aiki tare da ƙaramin farashi da rikitarwa, yana ba mutane da dama damar samun damar yin amfani da fasahohin aiki mai girma don nazarin bayanai.

Godiya ga goyan bayan GPUs, MapD shima yana ba da yanayi don ganin bayanan bayanai ta amfani da damar zane-zane na GPUs. Yana sauƙaƙe ƙirƙirar zane-zane mai ma'amala tare da babban adadin bayanai, yana ba da damar hulɗa tare da bayanin kusan a ainihin lokacin (maƙasudin mafarki na kowane mai nazarin bayanai). Toari da haɗawa da wasu algorithms na koyon inji (Ilmantarwa Na'ura), don yin ci gaba na bincike tare da yanayi iri ɗaya ta amfani da GPUs.

Muna gayyatarku zuwa yawo cikin Shafin hukuma na MapD don sake duba kowane fasalinsa daki-daki. Suna kuma bayar da takarda, wanda zaku iya zazzagewa kyauta, tare da yin cikakken bayani game da fasahohi da hanyoyin da suka sa MapD ya yiwu. Kuna iya jin daɗin wasu Demos abin mamaki!
MapD a halin yanzu yana cikin beta kuma akwai don Linux, zaku iya rubuta musu (tare da bayanin bayani) don shiga ciki.