gagarumin saki na tsarin sarrafa taron da aka rarraba , sananne don sauyawa zuwa sabon gine-ginen da aka aiwatar a Java, maimakon harshen Clojure da aka yi amfani da shi a baya.
Aikin yana ba ku damar tsara garantin sarrafa abubuwan da suka faru daban-daban a cikin ainihin lokaci. Misali, ana iya amfani da Storm don nazarin rafukan bayanai a cikin ainihin lokaci, gudanar da ayyukan koyon injin, tsara ƙididdiga mai ci gaba, aiwatar da RPC, ETL, da sauransu. Tsarin yana goyan bayan tari, ƙirƙirar saiti masu jurewa kuskure, yanayin sarrafa bayanai da garanti kuma yana da babban aiki, wanda ya isa aiwatar da buƙatun sama da miliyan ɗaya a sakan daya akan kullin tari.
Ana tallafawa haɗin kai tare da tsarin sarrafa jerin gwano iri-iri da fasahar bayanai. Tsarin gine-ginen Storm ya ƙunshi karɓa da sarrafa abubuwan da ba a tsara su ba, koyaushe ana sabunta rafukan bayanai ta amfani da na'urori masu sarƙoƙi na sabani tare da ikon rarraba tsakanin matakai daban-daban na lissafin. An mika aikin ga al'ummar Apache bayan Twitter ya mallaki BackType, kamfanin da ya kirkiro tsarin. A aikace, an yi amfani da Storm a BackType don nazarin tunanin abubuwan da suka faru a cikin microblogs, ta hanyar kwatanta kan tashi da sababbin tweets da kuma hanyoyin da aka yi amfani da su (alal misali, an kimanta yadda hanyoyin haɗin waje ko sanarwar da aka buga a kan Twitter aka sake watsawa ta wasu mahalarta. ).
Ana kwatanta ayyukan guguwa da dandamalin Hadoop, tare da babban bambanci shine cewa ba a adana bayanan a cikin ma'ajiyar ajiya, amma ana shigar da su a waje kuma ana sarrafa su cikin ainihin lokaci. Guguwa ba ta da ginanniyar ajiya mai ginanni kuma ana fara amfani da tambayar nazari kan bayanan da ke shigowa har sai an soke ta (yayin da Hadoop ke amfani da MapReduce jobs na iyakacin lokaci, Storm yana amfani da ra'ayin ci gaba da gudana "topologies"). Ana iya rarraba aiwatar da masu sarrafa su a cikin sabar da yawa - Storm yana daidaita aiki ta atomatik tare da zaren akan nodes daban-daban.
An fara rubuta tsarin a cikin Clojure kuma yana gudana a cikin na'ura mai mahimmanci na JVM. Gidauniyar Apache ta ƙaddamar da wani yunƙuri don ƙaura Storm zuwa sabon kwaya da aka rubuta a cikin Java, wanda aka gabatar da sakamakonsa a cikin sakin Apache Storm 2.0. Ana sake rubuta duk ainihin abubuwan da ke cikin dandalin a cikin Java. An ci gaba da tallafawa masu sarrafa rubutu a cikin Clojure, amma yanzu ana ba da su ta hanyar ɗaure. Storm 2.0.0 yana buƙatar Java 8. An sake fasalin ƙirar sarrafa zaren da yawa gaba ɗaya, yana barin karuwa mai gani a cikin aiki (ga wasu topologies, an rage jinkiri da 50-80%).
Sabuwar sigar kuma tana ba da sabon nau'in Streams API wanda ke ba ku damar ayyana masu sarrafa ta yin amfani da ayyukan salon shirye-shirye. Ana aiwatar da sabon API a saman daidaitaccen tushe API kuma yana goyan bayan haɗa kai tsaye na ayyuka don haɓaka sarrafa su. API ɗin Windowing don ayyukan taga ya ƙara goyan baya don adanawa da maidowa jihar a bayan baya.
Taimako don yin la'akari da ƙarin albarkatu lokacin yin yanke shawara ba'a iyakance ga
CPU da ƙwaƙwalwar ajiya, kamar cibiyar sadarwa da saitunan GPU. An yi babban adadin haɓakawa don tabbatar da haɗin kai tare da dandamali . An faɗaɗa tsarin kula da damar shiga don haɗawa da ikon ƙirƙirar ƙungiyoyi masu gudanarwa da wakilai. Ƙara haɓakawa masu alaƙa da SQL da tallafin awo. Sabbin umarni sun bayyana a cikin mahallin mai gudanarwa don gyara yanayin tari.
Yankunan aikace-aikace don Storm:
- Gudanar da rafukan sabbin bayanai ko sabunta bayanai a cikin ainihin lokaci;
- Ci gaba da Kwamfuta: Guguwa na iya gudanar da tambayoyin ci gaba da aiwatar da rafuka masu ci gaba, yana ba da sakamakon sarrafawa ga abokin ciniki a ainihin lokacin.
- Kiran Tsarin Nesa Rarraba (RPC): Ana iya amfani da guguwa don samar da aiwatar da tambayoyi masu ƙarfi a layi daya. Aiki ("topology") a cikin Storm aiki ne da aka rarraba a ko'ina da ke jiran saƙon ya zo waɗanda ake buƙatar sarrafawa. Bayan karɓar saƙo, aikin yana sarrafa shi a cikin mahallin gida kuma ya dawo da sakamakon. Misali na yin amfani da RPC da aka rarraba zai kasance don aiwatar da tambayoyin bincike a layi daya ko yin ayyuka akan babban saiti.
Siffofin guguwa:
- Samfurin shirye-shirye mai sauƙi wanda ke sauƙaƙe sarrafa bayanai na lokaci-lokaci;
- Taimako ga kowane yarukan shirye-shirye. Modules suna samuwa don Java, Ruby da Python, daidaitawa ga wasu harsuna yana da sauƙi godiya ga tsarin sadarwa mai sauƙi wanda ke buƙatar kusan layin 100 don tallafawa;
- Haƙuri kuskure: don gudanar da aikin sarrafa bayanai, kuna buƙatar ƙirƙirar fayil ɗin jar tare da lamba. Guguwa za ta rarraba wannan babban fayil ɗin gwal a kan kuɗaɗen tari, ya haɗa masu kula da shi, da tsara sa ido. Lokacin da aikin ya cika, za a kashe lambar ta atomatik akan duk nodes;
- Daidaitaccen scalability. Ana yin duk lissafin a cikin yanayin layi ɗaya; yayin da nauyin ke ƙaruwa, ya isa kawai a haɗa sabbin nodes zuwa gungu;
- Abin dogaro. Guguwa yana tabbatar da cewa kowane saƙo mai shigowa ana sarrafa shi aƙalla sau ɗaya. Za a sarrafa saƙon sau ɗaya kawai idan babu kurakurai yayin wucewa ta duk masu sarrafa; idan matsaloli suka taso, to za a sake yin ƙoƙarin sarrafa abin da bai yi nasara ba.
- Gudu. An rubuta lambar Storm tare da babban aiki a zuciya kuma yana amfani da tsarin don saurin saƙon asynchronous .
source: budenet.ru