Za a gudanar da taro a ranar 11-12 ga Yuli a St. Petersburg , sadaukar don haɓaka tsarin layi daya da rarrabawa. Dabarar Hydra ita ce ta haɗa ƙwararrun masana kimiyya (waɗanda galibi ana iya samun su kawai a taron kimiyya na ƙasashen waje) da mashahuran injiniyoyi masu aiki a cikin babban shiri guda ɗaya a mahadar kimiyya da aiki.
Hydra na ɗaya daga cikin muhimman tarukanmu a cikin ƴan shekarun da suka gabata. An riga an yi shiri sosai, zaɓin masu magana da rahotanni. Makon da ya gabata game da wannan tare da darektan JUG.ru Group, Alexey Fedorov ().
Mu game da muhimman mahalarta uku, waɗanda suka kafa ka'idar tsarin rarraba - Leslie Lamport, Maurice Herlihy da Michael Scott. Lokaci ya yi da za a yi magana dalla-dalla game da dukan shirin!
Motsawa
Idan kana da hannu a cikin shirye-shirye, to wata hanya ko wata kana mu'amala da multithreading da rarraba computing. Masana a cikin filayen da suka dace suna aiki tare da su kai tsaye, amma a fakaice, rarrabawa yana kallon mu daga ko'ina: a cikin kowace kwamfuta mai mahimmanci ko sabis na rarraba akwai wani abu da ke yin lissafin a layi daya.
Akwai tarurruka da yawa waɗanda suka shafi fannoni daban-daban na shirye-shiryen aikace-aikacen. A gefe guda na bakan, muna da ƙwararrun makarantun kimiyya waɗanda ke bayyana ɗimbin ka'idoji masu rikitarwa a cikin tsarin lacca. Alal misali, a cikin layi daya da Hydra a St. Petersburg akwai . A taron Hydra, mun yi ƙoƙari mu haɗu da muggan ayyuka, kimiyya, da duk abin da ke tsakanin su.
Ka yi tunani game da wannan: muna rayuwa a cikin wani lokaci mai ban mamaki lokacin da za ku iya saduwa da waɗanda suka kafa fannin kimiyya da injiniya da muke nazari. Masana kimiyya ba za su hadu da Newton ko Einstein ba - jirgin kasa ya tafi. Amma kusa da mu har yanzu suna rayuwa waɗanda suka ƙirƙiri tushen ka'idar rarraba tsarin, ƙirƙira shahararrun harsunan shirye-shirye, kuma a karon farko sun ƙunshi duk wannan a cikin samfuran aiki. Wadannan mutane ba su bar ayyukansu ba, a halin yanzu suna aiki kan batutuwa masu mahimmanci a manyan jami'o'i da kamfanoni na duniya, kuma sune mafi girma tushen ilimi da kwarewa a yau.
A daya hannun, da damar saduwa da su yawanci ya kasance zalla ka'idar: kadan daga cikin mu iya kullum saka idanu jama'a al'amurran da suka shafi a wasu Jami'ar Rochester, sa'an nan kuma garzaya zuwa Amurka da kuma mayar da wani lacca tare da Michael Scott. Ziyartar duk membobin Hydra zai kashe kuɗi kaɗan, ba tare da ƙidayar abyss na ɓata lokaci ba (ko da yake yana kama da nema mai ban sha'awa).
A gefe guda, muna da manyan injiniyoyi da yawa waɗanda ke aiki akan matsalolin matsa lamba a cikin tsarin rarrabawa a yanzu, kuma tabbas suna da abubuwa da yawa don faɗa. Amma ga matsalar - su .аботают, kuma lokacinsu yana da daraja. Haka ne, idan kai ma'aikaci ne na Microsoft, Google ko JetBrains, yiwuwar saduwa da ɗaya daga cikin mashahuran masu magana a wani taron ciki yana ƙaruwa sosai, amma a gaba ɗaya, a'a, wannan ba ya faruwa kowace rana.
Ta wannan hanyar, taron Hydra yana aiwatar da wani muhimmin aiki wanda yawancinmu ba za su iya yi da kanmu ba - a wuri ɗaya kuma a lokaci guda, yana haɗa mutanen da ra'ayoyinsu ko hulɗarsu da za su iya canza rayuwar ku. Na yarda cewa ba kowa ba ne ke buƙatar tsarin rarraba ko wasu abubuwa masu mahimmanci. Kuna iya tsara CRUDs a cikin PHP har tsawon rayuwar ku kuma ku kasance cikin farin ciki gaba ɗaya. Amma duk wanda yake bukata, wannan shine damar ku.
Tsawon lokaci ya wuce tun farkon sanarwar taron Hydra akan Habré. A wannan lokacin, an yi ayyuka da yawa - kuma yanzu muna da jerin sunayen kusan dukkanin rahotanni. Babu algorithms masu zare guda ɗaya masu sluggis, kawai tsattsauran hardcore da aka rarraba! Bari mu gama da kalmomi gaba ɗaya mu ga abin da ke hannunmu yanzu.
Mahimman bayanai
Mahimman bayanai sun fara da ƙare kwanakin taron. Yawancin lokaci maƙasudin mabuɗin buɗewa shine saita ruhi da alkiblar taron. Mahimmin bayanin rufewa ya zana layi kuma ya bayyana yadda za mu iya rayuwa tare da ilimi da basirar da aka samu yayin taron. Farko da ƙarshe: abin da aka fi tunawa da shi, kuma a gaba ɗaya, ya karu da mahimmanci.
Cliff Danna
Cliff labari ne a duniyar Java. A ƙarshen 90s, don karatun digiri na PhD, ya rubuta takarda mai suna , wanda daga baya ya zama tushen HotSpot JVM Server Compiler. Shekaru biyu bayan haka, ya riga ya yi aiki a Sun Microsystems akan JVM kuma ya nuna duk duniya cewa JIT yana da hakkin ya wanzu. Wannan labarin gaba ɗaya game da yadda Java ke ɗaya daga cikin mafi saurin gudu na zamani tare da mafi wayo da haɓakawa mafi sauri ya fito daga Cliff Click. A farkon farkon, an yi imani cewa idan wani abu yana iya samun dama ga mai tarawa a tsaye, ba lallai ne ku yi ƙoƙari ku jit ba. Godiya ga aikin Cliff da ƙungiyar, duk sabbin harsuna sun fara ƙirƙirar tare da ra'ayin tattara JIT ta tsohuwa. Tabbas, wannan ba aikin mutum ɗaya ba ne, amma Cliff ya taka muhimmiyar rawa a cikinsa.
A cikin maɓallin buɗewa, Cliff zai yi magana game da sauran ƙoƙarinsa - , wani dandamali na ƙwaƙwalwar ajiya don rarrabawa da ƙirar injin koyo don aikace-aikacen masana'antu. Ko fiye daidai, game da rarraba ma'auni na maɓalli-darajar nau'i-nau'i a ciki. Wannan babban ajiya ne mai sauri tare da kaddarorin ban sha'awa da yawa (ainihin lissafin yana ciki ), wanda ke ba da damar yin amfani da irin wannan mafita a cikin ilimin lissafi na manyan bayanai.
Wani rahoton da Cliff zai bayar shine - . Wani bangare na tarihin rayuwarsa - shekaru goma , Inda ya sabunta kuma ya inganta abubuwa da yawa a cikin kayan aikin Azul da fasaha na fasaha: JIT compilers, runtime, model model, kuskuren kulawa, sarrafa kaya, katsewar hardware, loading class, da sauransu da sauransu - da kyau, kuna samun ra'ayi.
Bangaren ban sha'awa ya fara ne lokacin da suka kera kayan masarufi don babban kasuwanci - supercomputer don gudanar da Java. Wani sabon abu ne mai ban sha'awa, wanda aka keɓance musamman don Java, wanda ke da buƙatu na musamman - karanta shingen ƙwaƙwalwar ajiya don tarin datti mai ƙarancin tsayawa, tsararru tare da bincika iyakoki, kiran kama-da-wane ... Ɗayan mafi kyawun fasahar shine ƙwaƙwalwar ajiyar kayan masarufi. Duk L1 na kowane nau'i na 864 na iya shiga cikin rubuce-rubucen ma'amala, wanda ke da mahimmanci musamman don aiki tare da makullai a cikin Java (tubalan da aka daidaita zasu iya aiki a layi daya muddin babu wani rikici na ƙwaƙwalwar ajiya na gaske). Amma kyakkyawan ra'ayin ya murkushe ta da mummunan gaskiya - kuma a cikin wannan magana Cliff zai gaya muku dalilin da yasa HTM da STM ba su dace da buƙatun ƙididdiga masu yawa ba.
Michael Scott -
- Farfesan Kimiyyar Kwamfuta a Jami'ar Rochester, wanda rabo ya haɗu da shi , kuma a gidansa na Jami'ar Wisconsin-Madison, ya kasance shugaban na tsawon shekaru biyar. Yana bincike da koyar da ɗalibai game da layi ɗaya da rarraba shirye-shirye da ƙirar harshe.
Duk duniya ta san Michael godiya ga littafin karatu , sabon bugu wanda aka buga kwanan nan - a cikin 2015. Aikin sa karɓa a matsayin daya daga cikin mafi shahara a fagen rarraba kwamfuta da a Jami'ar Rochester Online Library. Kuna iya sanin shi a matsayin marubucin ainihin Michael-Scott algorithm daga .
Dangane da duniyar Java, wannan lamari ne na musamman: tare da Doug Lea, ya haɓaka algorithms marasa toshewa da layukan daidaitawa waɗanda ɗakunan karatu na Java ke aiki. Wannan shi ne ainihin abin da mahimmin bayanin "Tsarin bayanan Dual" zai kasance game da shi - gabatarwar waɗannan tsarin a cikin Java SE 6 ya inganta aikin da sau 10. java.util.concurrent.ThreadPoolExecutor. Idan kuna mamakin abin da waɗannan "Tsarin bayanan Dual" suke, to akwai bayanai game da shi .
Maurice Herlihy -
- wanda ya lashe kyautar Dijkstra guda biyu. Na farko don aiki ne (Jami'ar Brown), da na biyu, mafi kwanan nan - (Jami'ar Fasaha ta Virginia). Kyautar Dijkstra ta gane aikin da muhimmancinsa da tasirinsa ya kasance a bayyane na akalla shekaru goma, kuma Maurice a fili yana daya daga cikin mashahuran masana a fagen. A halin yanzu yana aiki a matsayin farfesa a Jami'ar Brown kuma yana da jerin abubuwan ci gaba mai tsayi.
A cikin wannan mahimmin bayanin rufewa, Maurice zai yi magana game da ka'idar da kuma aiwatar da tsarin rarraba blockchain daga ra'ayi na litattafan ƙididdiga na ƙididdigar rarraba da kuma yadda yake sauƙaƙa yawancin matsalolin da ke da alaƙa. Wannan rahoto ne na musamman kan batun taron - ko kaɗan ba game da haƙar ma'adinai ba, sai dai game da yadda za a iya amfani da iliminmu da ban mamaki yadda ya kamata da kuma dacewa dangane da ayyuka iri-iri.
A cikin Yuli 2017, Maurice ya riga ya zo Rasha don halartar makarantar SPTDC, ya halarci taron JUG.ru, kuma ana iya kallon rikodin akan YouTube:
Babban shirin
A gaba za a yi takaitaccen bayani kan rahotannin da ke cikin shirin. An yi bayanin wasu daga cikin rahotannin a nan dalla-dalla, wasu kuma a takaice. Dogayen kwatance sun tafi galibi zuwa rahotannin harshen Ingilishi waɗanda ke buƙatar haɗin kai zuwa takaddun kimiyya, sharuɗɗan akan Wikipedia, da sauransu. Ana samun cikakken lissafin . Za a sabunta lissafin da ke kan gidan yanar gizon kuma a ƙara su.
Leslie Lamport -
Leslie Lamport ita ce mawallafin ayyukan seminal a cikin rarraba kwamfuta. yana nufin "Lamport TeX". Shi ne wanda ya fara, a cikin 1979, ya gabatar da manufar , da labarinsa ya sami lambar yabo ta Dijkstra.
Wannan shi ne bangaren da ba a saba gani a cikin shirin ba ta fuskar tsari, domin ba ma rahoto ba ne, sai dai na tambayoyi da amsa. Lokacin da wani muhimmin ɓangare na masu sauraro ya riga ya saba (ko zai iya zama saba) tare da kowane nau'i na ayyuka bisa "ka'idar Lamport", labaransa da rahotanni, yana da mahimmanci don ciyar da duk lokacin da ake samuwa akan sadarwa kai tsaye.
Tunanin yana da sauƙi - kuna kallon rahotanni guda biyu akan YouTube: и kuma shirya aƙalla tambaya ɗaya, kuma Leslie ta amsa.
Na farko daga cikin waɗannan bidiyoyi guda biyu da muke da su . Idan ba ku da sa'a guda don kallon bidiyon, zaku iya karanta shi da sauri a cikin sigar rubutu.
Lura: Akwai ƙarin bidiyoyi na Leslie Lamport akan YouTube. Misali, akwai mai girma . Ana samun sigar layi na gabaɗayan wannan kwas a , kuma ya loda shi zuwa YouTube don sauƙin dubawa akan na'urorin hannu.
Martin Kleppmann -
Martin Kleppmann mai bincike ne a Jami'ar Cambridge yana aiki akan CRDT da kuma tabbatar da al'adar algorithms. Littafin Martin , wanda aka buga a cikin 2017, ya tabbatar da samun nasara sosai kuma ya sanya shi cikin jerin masu siyarwa a fagen adana bayanai da sarrafa bayanai. Kevin Scott, CTO na Microsoft, : “Ya kamata wannan littafi ya zama dole ga injiniyoyin software. Wannan wata hanya ce da ba kasafai ba wacce ta haɗu da ka'ida da aiki don taimakawa masu haɓakawa da hankali wajen ƙira da aiwatar da abubuwan more rayuwa da tsarin bayanai." Mahaliccin Kafka da CTO na Confluent, Jay Kreps, ya faɗi wani abu makamancin haka.
Kafin ya shiga cikin binciken ilimi, Martin ya yi aiki a masana'antu kuma ya kafa ƙungiyoyi biyu masu nasara:
- Rapportive, sadaukar don nuna bayanan zamantakewa na lambobin sadarwa daga imel ɗin ku, wanda LinkedIn ya saya a cikin 2012;
- Go Test It, sabis don gwada gidajen yanar gizo ta atomatik a cikin mazugi daban-daban, wanda RedGate ya saya a 2009.
Gabaɗaya, Martin, ko da yake bai shahara fiye da jigogin mu ba, ya riga ya sami damar ba da gudummawa ga haɓaka ƙididdigar rarrabawa da masana'antu.
A cikin wannan jawabin, Martin zai yi magana game da wani batu kusa da bincikensa na ilimi. A cikin Google Docs da makamantansu na gyara sofas, "gyare-gyaren haɗin gwiwa" yana nufin aikin kwafi: kowane mai amfani yana da nasa kwafi na takaddun da aka raba, wanda sai su gyara, kuma ana aika duk canje-canje a cikin hanyar sadarwa zuwa sauran abubuwan. mahalarta. Canje-canje ga takaddun layi ba tare da layi ba suna haifar da rashin daidaituwa na ɗan lokaci na takaddar dangane da sauran mahalarta, kuma sake daidaitawa yana buƙatar sarrafa rikici. Wannan shine ainihin abin da suke wanzuwa (CRDT), a gaskiya, sabon abu ne mai gaskiya, wanda aka tsara shi kawai a cikin 2011. Wannan magana ta tattauna abin da ya faru tun lokacin a cikin duniyar CRDT, menene ci gaba na baya-bayan nan, hanyar ƙirƙirar aikace-aikacen gida-farko gabaɗaya da kuma amfani da ɗakin karatu mai buɗewa. musamman.
Mako mai zuwa za mu buga doguwar hira da Martin akan Habré, zai kasance mai ban sha'awa.
Pedro Ramalhete -
Pedro yana aiki a Sisiko kuma yana haɓaka algorithms na layi ɗaya na shekaru goma da suka gabata ko makamancin haka, gami da hanyoyin aiki tare, tsarin bayanai mara-kulle da mara jira da duk abin da zaku iya tunanin akan wannan batu. Bincikensa na yanzu da buƙatun aikin injiniya yana mai da hankali kan Ginin Universal, Ƙwaƙwalwar Ma'amalar Software, Ƙwaƙwalwar Ƙwaƙwalwar Ƙwaƙwalwa da makamantansu waɗanda ke ba da damar aikace-aikacen daidai, daidaitawa da kuskure. Shi ne mawallafin bulogin da aka fi sani a cikin kunkuntar da'ira .
Yawancin aikace-aikace masu zare da yawa yanzu suna gudana akan tsarin bayanai iri ɗaya, tun daga amfani da layin saƙo tsakanin 'yan wasan kwaikwayo zuwa tsarin tsarin bayanai a cikin shagunan ƙima. Suna aiki cikin nasara a cikin Java JDK shekaru da yawa, kuma a hankali ana ƙara su zuwa C++.
Hanya mafi sauƙi don aiwatar da tsarin bayanai na layi ɗaya shine aiwatarwa na jeri (mai zare guda ɗaya) wanda hanyoyin da ake kiyaye su ta hanyar ɓata. Wannan yana da damar zuwa kowane Yuni, amma yana da matsaloli na zahiri tare da ƙima da aiki. A lokaci guda, tsarin bayanan da ba a kulle ba da jira ba wai kawai ya fi dacewa da kurakurai ba, har ma suna da ingantaccen bayanin martaba - duk da haka, ci gaban su yana buƙatar ƙwarewa mai zurfi da daidaitawa ga takamaiman aikace-aikacen. Ɗayan layin kuskure ɗaya ya isa ya karya komai.
Ta yaya za mu iya yin shi ta yadda ko da wanda ba ƙwararre ba zai iya tsarawa da aiwatar da irin waɗannan tsarin bayanai? An sani cewa duk wani jerin algorithm za a iya sanya zaren lafiya ta amfani da ko dai , ko ƙwaƙwalwar ajiyar ma'amala. Abu ɗaya, za su iya rage shingen shiga cikin magance wannan matsalar. Koyaya, duka mafita yawanci suna haifar da aiwatarwa mara inganci. Pedro zai yi magana game da yadda suka gudanar da yin waɗannan ƙira mafi inganci da kuma yadda za ku iya amfani da su don algorithms.
Heidi Howard -
Heidi Howard shine, kamar Martin, mai binciken tsarin rarrabawa a Jami'ar Cambridge. Kwarewarta sune daidaito, rashin haƙuri, aiki da ra'ayi mai rarraba. An fi saninta da ta gama-gari na Paxos algorithm da ake kira .
Ka tuna cewa dangi ne na ka'idoji don magance matsalar yarjejeniya a cikin hanyar sadarwa na kwamfutoci marasa aminci, dangane da aikin Leslie Lamport. Don haka, wasu daga cikin masu magana da mu suna aiki kan matsalolin da wasu masu magana da mu suka gabatar da su tun asali - kuma wannan yana da ban mamaki.
Ikon samun yarjejeniya tsakanin runduna da yawa-don magancewa, zaɓen shugabanni, toshewa, ko haɗin kai-shine muhimmin batu a cikin tsarin rarrabawar zamani. Paxos yanzu shine babbar hanyar magance matsalolin yarjejeniya, kuma akwai bincike da yawa da ke gudana a kusa da shi don faɗaɗawa da haɓaka algorithm don buƙatu masu amfani daban-daban.
A cikin wannan magana, za mu sake duba tushen ka'idar Paxos, shakatawa da buƙatun asali da haɓaka algorithm. Za mu ga cewa Paxos ainihin zaɓi ɗaya ne kawai a tsakanin ɗimbin hanyoyin haɗin gwiwa, kuma sauran abubuwan da ke kan bakan suna da amfani sosai don gina ingantaccen tsarin rarrabawa.
Alex Petrov -
Alex kwararre ne na bayanai da tsarin ajiya, kuma mafi mahimmanci a gare mu, mai aiwatarwa a ciki . A halin yanzu yana aiki akan littafi, Database Internals, tare da O'Reilly.
Don tsarin tare da (a cikin kalmomin Rasha - "madaidaicin daidaito"), bayan kumburin kumburi ko rarrabuwar hanyar sadarwa, kuna buƙatar warware matsalar mai zuwa: ko dai ci gaba da aiwatar da buƙatun, sadaukar da daidaito, ko ƙi aiwatar da su da sadaukarwa samuwa. A cikin irin wannan tsarin, ƙididdiga, rukunoni masu ruɓani na nodes da kuma tabbatar da cewa aƙalla kumburi ɗaya ya ƙunshi ƙima na baya-bayan nan, na iya zama mafita mai kyau. Kuna iya tsira da gazawa da asarar haɗin kai zuwa wasu kumburi yayin da kuke amsawa da sabbin ƙima.
Koyaya, komai yana da farashin sa. Tsarin kwafin ƙididdiga yana nufin ƙarin farashi na ajiya: dole ne a adana bayanan da ba su da yawa a kan nodes da yawa lokaci ɗaya don tabbatar da cewa akwai isassun kwafi lokacin da matsala ta faru. Ya zama cewa ba dole ba ne ka adana duk bayanan akan duk kwafi. Kuna iya rage nauyi akan ma'ajiyar idan kun adana bayanai kawai akan ɓangaren nodes, kuma kuyi amfani da nodes na musamman (Transient Replica) don gazawar sarrafa yanayin.
A yayin gudanar da rahoton za mu yi la'akari , tsarin kwafi da aka yi amfani da shi a ciki и , da aiwatar da wannan ra'ayi a cikin Apache Cassandra da ake kira .
Dmitry Vyukov -
Dmitry mai haɓakawa ne a Google yana aiki akan gwaji mai ƙarfi don C/C++ da Go - Address/Memory/ThreadSanitizer, da makamantan kayan aikin Linux kernel. An ba da gudummawa don Tafi mai tsara jadawalin goroutine, mai jefa kuri'a na hanyar sadarwa, da mai tara shara iri ɗaya. Shi kwararre ne a cikin multithreading, marubucin dozin sabbin algorithms marasa toshewa kuma shine mai Intel.
Yanzu kadan game da rahoton da kansa. Harshen Go yana da goyon baya na asali don multithreading a cikin nau'i na goroutines (zaren haske) da tashoshi (layin FIFO). Wadannan hanyoyin suna sa ya zama mai sauƙi da jin daɗi ga masu amfani don rubuta aikace-aikacen zamani masu yawa, kuma yana kama da sihiri. Kamar yadda muka fahimta, babu wani sihiri a nan. A cikin wannan magana, Dmitry zai shiga cikin abubuwan da ke tattare da tsarin tsarin Go kuma ya nuna asirin aiwatar da wannan "sihiri". Da farko, zai ba da taƙaitaccen bayani game da manyan abubuwan da ke cikin jadawalin kuma ya gaya muku yadda yake aiki. Na gaba, za mu yi la'akari sosai kan fannonin ɗaiɗaikun kamar dabarun ajiye motoci da ajiye fakin da kuma yadda ake toshe kiran tsarin. A ƙarshe, Dmitry zai yi magana kaɗan game da yiwuwar ingantawa ga mai tsarawa.
Dmitry Bugaichenko -
Dmitry yayi aiki a cikin fitar da kayayyaki kusan shekaru 9 ba tare da rasa alaƙa da jami'a da al'ummar kimiyya ba. Babban bincike na bayanai a Odnoklassniki ya zama masa wata dama ta musamman don haɗa horo na ka'idar da tushe na kimiyya tare da haɓaka samfuran gaske, waɗanda ake buƙata.
Binciken jadawali da aka rarraba ya kasance kuma ya kasance aiki mai wuyar gaske: lokacin da ya zama dole don samun bayanai game da haɗin kai na maƙwabcin maƙwabta, sau da yawa dole ne a canja wurin bayanai tsakanin na'urori, wanda ke haifar da ƙarin lokacin kisa da kaya a kan kayan aikin cibiyar sadarwa. A cikin wannan magana, za mu ga yadda za ku iya samun saurin sarrafawa ta hanyar amfani da tsarin bayanai masu yuwuwa ko hujjoji kamar ma'auni na jadawalin abokantaka a cikin hanyar sadarwar zamantakewa. Duk waɗannan ana kwatanta su da misalan lamba a cikin Apache Spark.
Denis Rystsov -
Denis - developer , ƙwararre a cikin duba samfuran daidaito, algorithms yarjejeniya, da ma'amaloli da aka rarraba. A halin yanzu yana aiki a Microsoft, kuma kafin wannan ya yi aiki akan tsarin rarrabawa a Amazon da Yandex.
A cikin wannan magana, za mu yi la'akari da ka'idojin ciniki da aka rarraba da aka ƙirƙira a cikin 'yan shekarun da suka gabata, waɗanda za a iya aiwatar da su a gefen abokin ciniki a saman kowane kantin sayar da bayanai wanda ke goyan bayan sabunta yanayin (kwatanta da saita). Maganar ƙasa ita ce, rayuwa ba ta ƙare da ƙaddamarwa na kashi biyu, ana iya ƙara ma'amaloli a saman kowane bayanan bayanai - a matakin aikace-aikacen, amma ka'idoji daban-daban (2PC, Percolator, RAMP) suna da ciniki daban-daban kuma ba a ba mu ba. kyauta.
Alexei Zinoviev -
Alexei () mai magana ne na dogon lokaci kuma memba na kwamitocin shirye-shirye a wasu tarukan. Koyarwa mai koyarwa a EPAM Systems, kuma ya kasance abokai tare da Hadoop/Spark da sauran manyan bayanai tun 2012.
A cikin wannan magana, Alexey zai yi magana game da matsalolin daidaita tsarin koyan injin na gargajiya don aiwatarwa a cikin yanayin rarraba bisa ga kwarewarsa tare da Apache Spark ML, Apache Mahout, Apache Flink ML da ƙwarewar ƙirƙirar Apache Ignite ML. Alexey kuma zai yi magana game da aiwatar da algorithms na ML da aka rarraba a cikin waɗannan tsarin.
Kuma a ƙarshe, rahotanni guda biyu daga Yandex game da Yandex Database.
Vladislav Kuznetsov -
Vladislav mai haɓakawa ne a Yandex a cikin rukunin dandamali da aka rarraba. Yanar Gizon Yanar Gizon Yanar Gizon Yanar Gizon Yanar Gizon Yanar Gizon Yanar Gizon Yanar Gizon Yanar Gizo, Rarraba Geo, DBMS mai jure rashin kuskure wanda zai iya jure gazawar fayafai, sabar, racks da cibiyoyin bayanai ba tare da rasa daidaito ba. Don tabbatar da haƙuri da kuskure, ana amfani da algorithm na mallakar mallaka don cimma yarjejeniya da aka rarraba, da kuma yawan hanyoyin fasaha, waɗanda aka tattauna dalla-dalla a cikin rahoton. Rahoton na iya zama mai ban sha'awa ga masu haɓaka DBMS da masu haɓaka hanyoyin magance aikace-aikacen bisa DBMS.
Semyon Checherinda -
Semyon shine mai haɓakawa a cikin rukunin dandamali da aka rarraba a Yandex, yana aiki akan yuwuwar amfani da masu haya da yawa na shigarwar YDB.
Database Yandex an tsara shi don tambayoyin OLTP kuma ya dace da buƙatun ACID don tsarin ma'amala. A cikin wannan rahoto, za mu yi la'akari da tsarin tsarin ma'amala wanda ke ƙarƙashin tsarin ma'amalar YDB. Bari mu dubi waɗanne ƙungiyoyi ne ke shiga cikin ma'amaloli, waɗanda ke ba da tsari na duniya ga ma'amaloli, yadda ake samun atomity na ma'amala, amintacce, da tsauraran matakin keɓewa. Yin amfani da matsala ta gama gari a matsayin misali, bari mu kalli aiwatar da ma'amala ta amfani da ayyuka biyu da ƙaddarar ma'amaloli. Mu tattauna sabaninsu.
Abin da ke gaba?
Ana ci gaba da cika shirin taron da sabbin rahotanni. Musamman, muna sa ran rahoto daga Nikita Koval () daga JetBrains da Oleg Anastasev () daga kamfanin Odnoklassniki. Nikita yana aiki akan algorithms don coroutines a cikin ƙungiyar Kotlin, kuma Oleg yana haɓaka gine-gine da mafita don tsarin ɗaukar nauyi a cikin dandalin Odnoklassniki. Bugu da ƙari, akwai ƙarin ramukan fanko na 1, kwamitin shirin yana aiki tare da 'yan takara a yanzu.
Taron Hydra zai gudana ne a ranar 11-12 ga Yuli a St. Petersburg. Akwai tikiti . Da fatan za a kula da samun tikitin kan layi - idan saboda wasu dalilai ba za ku iya zuwa St. Petersburg kwanakin nan ba.
Mu hadu a Hydra!
source: www.habr.com