Namuhla sibhekene nesimo se Idatha enkulu, singathola inani elikhulu lemininingwane kusuka kunombolo engapheli yemithombo. Le datha enkulu kangaka iletha izinzuzo eziningi, kepha futhi iletha izinselelo eziningi. Okuvame kakhulu kubo: Izikhathi zokuphendula kudathasethi yenqwaba.
ImephuD wazalelwa ukunikela ngesivinini esikhulu emkhakheni wolwazi oluhlaziywayo. Idizayinelwe ukucubungula izigidigidi zamarekhodi endabeni yama-millisecond sisebenzisa amandla wekhompyutha ahlinzekwa ngu- Ama-GPU. Yakhelwe kahle ukusebenzisa ngokugcwele wonke amakhono wehardware nesoftware atholakala kumakhadi emidwebo, inikeza abahlaziyi nezikhathi zokuphendula kososayensi bemininingwane mayelana nama-oda ama-3 ubukhulu (x1000) ngaphezulu kobuchwepheshe obusetshenziselwe lezi zinhloso ngaphambili. Ukusizakala ngokufana kwama-GPU (Cishe ama-cores angama-80000 kuma-GPU anamuhla) namabhendi we-memory memory amakhulu (Around 8Gbps) ukwenza i-algebra eqondile nokusesha kwe-database, usebenzisa i-LLVM ukuhlanganisa ngesikhathi sangempela ngasinye ukubonisana, ngaphezu kokugcina idatha ebonwe kakhulu kunqolobane yama-GPU (izinkumbulo ezisezingeni eliphezulu ze-DDR5).
Kufanele sikhumbule ukuthi emhlabeni we-Big Data, imininingwane yolwazi yendabuko ayisetshenziswa, ngokuya ngokubhalwa nokulondolozwa kwamafayela, ngoba lokhu kungadala inani eleqile lemisebenzi ye-I / O kwidiski enzima. Ngenhloso yokuhlaziya izigidigidi zamarekhodi, i- yolwazi lwenkumbulo, njengo-Apache Spark. Kodwa-ke, ukuthola inani lememori elidingekayo nokusebenza okufisayo, udinga iqoqo leseva futhi siyazi ukuthi lokhu kusho izindleko kwi-hardware, ukufakwa kwenethiwekhi nenani elikhulu labachwepheshe. Ngakho-ke, ImephuD inikeza amandla okufeza ukusebenza okuphezulu ngezindleko eziphansi nobunzima, okuvumela abantu abaningi ukuthi bakwazi ukufinyelela kubuchwepheshe obuphezulu bokusebenza kokuhlaziywa kwedatha.
Ngenxa yokuxhaswa yi-GPUs, iMapD nayo inikeza indawo yokubukwa kwedatha isebenzisa amandla wehluzo ama-GPU. Kusiza ukwakhiwa kwamagrafu asebenzisanayo ngevolumu ephezulu yedatha, okuvumela ukuxhumana nolwazi cishe ngesikhathi sangempela (iphupho elimanzi labo bonke abahlaziyi bemininingwane). Ngaphezu kokufaka amanye ama-algorithm wokufunda ngomshini (i-Machine Learning), ukwenza ukuhlaziywa okuthuthukile ngemvelo efanayo usebenzisa ama-GPU.
Sikumema ukuba uzulazule ku Ikhasi elisemthethweni leMapD ukubuyekeza ngayinye yezici zayo ngokuningiliziwe. Banikela nangephepha, ongalilanda mahhala, elinemininingwane yobuchwepheshe nezindlela ezenze iMapD yenzeka. Ungajabulela ezinye ama-demos Kuyamangaza!
IMapD okwamanje iku-beta futhi itholakalela i-Linux, ungababhalela (kanye nesitatimende esichazayo) ukuze ubambe iqhaza kuso.
Ungalokothi ucabange lolo hlobo lwento, uma ekuqaleni bekubonakala kungajwayelekile kimi ukucabanga kabusha, konke kwenzelwa kusengaphambili