Micron Technology, DRAM da kamfanin ƙwaƙwalwar ajiyar flash, sabon injin ajiya (Injin Ma'ajiyar Ƙwaƙwalwar Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwaƙwalwa ) na Ƙaddamarwa na Ƙaddamarwa na Ƙaddamarwa na Ƙaddamarwa da aka yi a kan faifan SSD dangane da NAND flash (X100, TLC, QLC 3D NAND) ko ƙwaƙwalwar ajiya na dindindin (NVDIMM). An ƙera injin ɗin azaman ɗakin karatu don haɗawa cikin wasu aikace-aikace kuma yana goyan bayan sarrafa bayanai a tsarin ƙima mai mahimmanci. An rubuta lambar HSE a cikin C da lasisi a ƙarƙashin Apache 2.0.
Daga cikin wuraren da ake amfani da injin ɗin, an ambaci ƙananan ma'ajin bayanai a cikin NoSQL DBMS, ma'ajiyar software (SDS, Ma'ajiyar Ƙa'idar Software) kamar Ceph da Scality RING, dandamali don sarrafa bayanai masu yawa (Big Data) , Babban tsarin sarrafa kwamfuta (HPC), na'urorin Intanet na abubuwa (IoT) da mafita don tsarin koyo na inji.
An inganta HSE ba kawai don iyakar aiki ba, har ma don tsawon rai a cikin nau'ikan azuzuwan SSD. Ana samun saurin aiki mai girma ta hanyar samfurin ajiya na matasan - mafi yawan bayanan da suka dace suna cache a cikin RAM, wanda ke rage yawan damar shiga motar. A matsayin misali na haɗa sabon injin zuwa ayyukan ɓangare na uku sigar DBMS MongoDB mai tushen daftarin aiki, wanda aka fassara don amfani da HSE.
A fasaha, HSE ya dogara da ƙarin ƙirar kwaya , wanda aiwatar da wani musamman abu ajiya dubawa ga m-jihar tafiyarwa, la'akari da su damar da fasali, wanda ba ka damar samun fundamentally daban-daban yi da karko halaye. Mpool kuma ci gaba ne na Fasahar Micron, wanda aka buɗe a lokaci guda da HSE, amma ya rabu cikin aikin samar da ababen more rayuwa mai zaman kansa. Mpool yana ɗaukar amfani и , amma a halin yanzu yana tallafawa SSDs na gargajiya kawai.
Gwajin aiki ta amfani da kunshin (Yahoo Cloud Serving Benchmark) ya nuna babban haɓakar aiki yayin amfani da ma'ajin tarin tarin fuka 2 tare da sarrafa tubalan bayanan 1 KB. Ana ganin haɓakar aiki mai mahimmanci musamman a cikin gwajin tare da rarraba iri ɗaya na ayyukan karantawa da rubutu (gwajin "A" a cikin jadawali).
Misali, MongoDB tare da injin HSE ya juya ya zama kusan sau 8 cikin sauri fiye da sigar tare da daidaitaccen injin WiredTiger, kuma RocksDB DBMS ya fi injin HSE sauri fiye da sau 6. Hakanan ana iya ganin kyakkyawan aiki a cikin gwaje-gwajen da suka ƙunshi ayyukan karanta 95% da 5% gyara ko ƙara ayyuka (gwaji "B" da "D" a cikin jadawali). Gwajin C, wanda ya ƙunshi ayyukan karantawa kawai, yana nuna riba kusan 40%. Haɓakawa na tsirar abubuwan tuƙi na SSD yayin ayyukan rubutu idan aka kwatanta da mafita dangane da RocksDB an kiyasta sau 7.
Babban fasali na HSE:
- Taimako ga ma'auni da kuma tsawaita masu aiki don sarrafa bayanai a cikin maɓalli / ƙima;
- Cikakken tallafi don ma'amaloli kuma tare da ikon keɓance sassan ajiya ta hanyar ƙirƙirar hotunan hoto (ana kuma iya amfani da hotunan hoto don kula da tarin masu zaman kansu a cikin ajiya ɗaya);
- Ikon yin amfani da siginan kwamfuta don karkatar da bayanai a cikin ra'ayi na tushen hoto;
- Samfurin bayanan da aka inganta don nau'ikan nau'ikan kaya masu gauraya a cikin ajiya guda;
- Hanyoyi masu sassauƙa don sarrafa amincin ajiya;
- Shirye-shiryen tsara bayanan ƙididdiga (rarraba a cikin nau'ikan ƙwaƙwalwar ajiya daban-daban waɗanda ke cikin ma'adana);
- Laburare tare da API C wanda zai iya haɗawa da kowane aikace-aikace;
- Ƙarfin ma'auni zuwa terabytes na bayanai da ɗaruruwan biliyoyin maɓalli a cikin ajiya;
- Ingantaccen aiki na dubban ayyuka na layi daya;
- Mahimmanci mai girma a cikin kayan aiki, rage jinkirin da ƙara yawan rubuce-rubuce / karantawa don nau'o'in nauyin aiki daban-daban idan aka kwatanta da daidaitattun hanyoyin mafita;
- Ikon yin amfani da faifan SSD na azuzuwan daban-daban a cikin ajiya ɗaya don haɓaka aiki da dorewa.
source: budenet.ru