Cache Layer saba bɛ Select ni Disk cɛ

Ni i ka baarakɛminɛn ye SELECT fɔcogo dɔ tasuma, o ɲininkali bɛ se ka maga disiki firifirili la abada walima hali flash marayɔrɔ raw — a bɛ tɛmɛ cakɛda saba danfaralenw fɛ minnu b’a dɔn u makuntɔ n’i ka jaabi bɛ se mikrosekɔndi walima milisekɔndi kɔnɔ. Nin layini ninnu faamuyali ye danfara ye jagokɛyɔrɔ min bɛ sɛgɛsɛgɛli kɛ k’a sɔrɔ cɛsiri ma kɛ, ani min bɛ siri diɲɛ doni lakika kɔrɔ.

SELECT ɲininkali bɛ bɔ i ka baarakɛminɛn na waati min na, mun bɛ kɛ ?

i ka baarakɛminɛn ye SELECT ɲininkali ci waati min na, a bɛ don pibiliki dɔ kɔnɔ, baarakɛlaw fanba tɛ min sɛgɛsɛgɛ abada. Donanw ka motɛri bɛ ɲinini in minɛ sani I/O si ka kɛ, ka SQL lajɛ ka kɛ kɔnɔna waleyali bolodalen ye ani k’a lafasali sira fɔlɔ lajɛ o yɔrɔnin bɛɛ: ɲininkali jaabiw marayɔrɔ. Ni ɲininkali kelen min bɛ ni paramɛtɛrɛ kelenw ye, o kɛra kɔsa in na, motɛri bɛ se ka jaabiw sigilen dɔ lasegin min maralen bɛ k’a sɔrɔ a ma maga kunnafonidilan ɲɛ kelen na. O bɛ wele tuma dɔw la ko ɲininkali cache walima jaabi cache, wa kalan caman, sɛbɛnni dɔgɔya baarakɛyɔrɔw kan — i n’a fɔ analytics dashboards ani reporting modules — a bɛ se ka disiki kalan fanba ban pewu.

hakilila kɔrɔlen min bɛ yan , o ye ko ɲininkalikɛyɔrɔ bɛ sɛgɛsɛgɛli kɛ kosɛbɛ kunnafonidilanw jiginni na . INSERT, UPDATE, walima DELETE o DELETE min bɛ tabali jukɔrɔ, o bɛ jaabi cakɛda minnu bɛ tali kɛ o la, olu tɛ tiɲɛ. O de y’a to sɛbɛnni-gɛlɛn jagokɛminɛnw ka teli ka ɲininkali cakɛda bali pewu, wa u bɛ u jigi da layɛrɛ jugumanw kan o nɔ na.

Buffer Pool ye mun ye ani mun na a nafa ka bon ka tɛmɛ i hakilina kan ?

Kasɛti layɛrɛ filanan — wa a bɛ se ka fɔ ko a nafa ka bon kosɛbɛ fɛn dilanni siratigɛ la — o ye buffer pool ye (a bɛ wele ko buffer shared PostgreSQL kɔnɔ, InnoDB buffer pool MySQL kɔnɔ). Nin ye RAM mara ye min bɛ Kɛ ni database engine ye walasa ka data ɲɛ minnu Sɔrɔla kɔsa in na olu Minɛ. Ni ɲininkali ma se ka kɛ ka bɔ jaabiw marayɔrɔ la, motɛri b’a lajɛ ni kunnafoni ɲɛ wajibiyalenw sigilen bɛ buffer pool kɔnɔ kaban sanni ka disiki kalanni si bɔ.

Buffer pool bɛ baara kɛ waati ni yɔrɔ sigida sariyakolo kan : kunnafonidilan minnu sɔrɔla kɔsa in na, a ka c’a la, olu bɛ se ka sɔrɔ kokura, wa kunnafonidilan minnu maralen bɛ kunnafonidilan sɔrɔlenw kɛrɛfɛ, a ka c’a la, olu bɛna sɔrɔ sɔɔni. Database ɲɛmɔgɔw bɛ buffer pool size tune i n’a fɔ u bɛ min kɛ ni configuration decisions hautes-leverage dɔ ye. Buffer pool min ka dɔgɔ kojugu, o bɛ kɛ sababu ye ka ɲɛw gɛnni kɛ tuma bɛɛ, ka ko dɔ lawuli min bɛ wele ko thrashing, o yɔrɔ la, sistɛmu bɛ waati caman kɛ ka cache misses ɲɛnabɔ ka tɛmɛ ɲininkaliw kɛli kan.

ye

Hakilila kunbaba : OLTP baarakɛta fanba la , buffer pool min bonya ka ɲi , o kɔrɔ ye ko kunnafonidilan kalanni bɛɛ 95–99% bɛ baara kɛ ka bɔ RAM la . Baarakɛjɛkulu — i ka kunnafonidilanw kulu fitinin min ɲininkaliw bɛ maga tiɲɛ na tuma caman na — a ka c’a la, o ka dɔgɔ kosɛbɛ ka tɛmɛ kunnafonidilan hakɛ bɛɛ kan. I ka buffer pool bonya ka bɛn i ka baarakɛjɛkulu ma, i kana i ka dataset bɛɛ ma, o ye tuning wale kelen ye min bɛ segin ka caya kosɛbɛ i bɛ se ka min kɛ.

Operasiyɔn ka cakɛda bɛ RAM ni Disiki cɛ danfara fa cogo di ?

Hali ni database yɛrɛ ka buffer pool ma se , ɲininkali ma bila fɔlɔ disiki kalanni lakika kama . Baarakɛminɛn bɛ page cache (a bɛ fɔ fana ko filesystem cache) mara, o ye RAM maralen ye min bɛ mara ni kernel ye, min bɛ kalanw ni sɛbɛnniw bufferw kɛ ka minɛnw bali. Ni database engine ye ɲɛ dɔ ɲini min tɛ a ka buffer pool kɔnɔ, OS kernel b’a yɛrɛ ka page cache lajɛ ka sɔrɔ ka I/O farikoloma cikan di storage controller ma.

Nin layini sabanan in fanba tɛ ye baarakɛminɛnw dilannikɛlaw fɛ nka a nafa ka bon kosɛbɛ sistɛmuw kan, kunnafonidilanw buffer pool ma labɛn yɔrɔ minnu na. OS ɲɛ cache bɛ tila-tila baarakɛcogo bɛɛ la, o la a bɛ ɲɔgɔn sɔrɔ ni i ka application server, web server ani porogaramu wɛrɛw ye minnu bɛ baara kɛ host kelen kan. Donanw ka sèrwɛri kɛrɛnkɛrɛnnenw kan, o ɲɔgɔndan in ka dɔgɔ, wa OS cache bɛ buffer filanan nafama di. Jatigila jɛlenw walima minɛn minnu hakilijagabɔ dan sirilen don, OS ka cakɛda ka teli ka dɔgɔya kojugu walasa ka dɛmɛ don.

Kasɛti layɛrɛ jumɛn de bɛ baara kɛcogo caman ɲɛnabɔ waleyali la ?

Diɲɛ lakika sɛnɛcogo la , buffer pool bɛ fanga sɔrɔ baarakɛcogo la ni danfaraba ye . Nin ye mun ye layini kelen-kelen bɛɛ bɛ dɛmɛ don cogo wɛrɛ la baarakɛcogo la:

• Ɲininkali jaabiw marayɔrɔ : Nafa caman bɛ kalan gɛlɛnw na, minnu fanba ye kunnafonidilanw ye minnu tɛ jɔ — ɲininkaliw fɔli, dakunw maralenw, foroba kɔnɔkow labanyɔrɔw. Nafa tɛ sɛbɛnni-gɛlɛn tabali kan.
• database buffer pool : diɲɛ bɛɛ ka baarakɛso . Production database server bɛɛ ka kan ka Labɛn yan fɔlɔ. A bɛ sɔrɔcogo gansanw ni sɔrɔcogo tugu-tugulenw bɛɛ ɲɛnabɔ ka ɲɛ.
• OS ɲɛ marayɔrɔ : A bɛ lakana sira di ni buffer pool hakɛ dɔgɔyara . A bɛ dɛmɛba fana kɛ tabali belebelew sɛgɛsɛgɛli tugu-tugulenw senfɛ minnu tun bɛna ɲɛ funtenimaw gɛn ka bɔ buffer pool kɔnɔ.
• Storage controller cache (hardware layer): Layer naaninan, min bɛ ye tuma caman na — NVMe SSDw ni RAID controllers bɛ sɛbɛnni cachew mara kurun kɔnɔ ni batiri walima kondansɛri kɔkanna ye. O bɛ muɲuli lakana k’a sɔrɔ a ma sɛbɛnni kɛcogo saraka fsync latency musaka la.
• Application-layer cache (Redis, Memcached): A sigilen bɛ kunnafonidilan sanfɛ pewu, a bɛ ɲininkali jaabiw walima jatebɔfɛnw mara walasa a kana kunnafonidilan gosi hali dɔɔnin — a ka ɲi kosɛbɛ SaaS plateformes multi-tenant minnu bɛ baara kɛ baarakɛla ba caman ye ɲɔgɔn fɛ.

Bi jagokɛyɔrɔw bɛ se ka baara kɛ ni Cache Architecture ye cogo di walasa ka dannaya sɔrɔ a hakɛ la ?

Jagokɛlaw fɛ minnu bɛ baara kɛ baarakɛda caman kɔnɔ — CRM, poroze ɲɛnabɔli, ɛntɛrinɛti jago, jateminɛ — cache architecture bɛ plateforme jaabi dantigɛ k’a ɲɛsin u ma ni ekipuw bɛ ka bonya. Plateforme minnu jɔlen bɛ cache fɛɛrɛ kan min bɛ layɛrɛ ɲuman na, olu bɛ se ka baara kɛ ni baarakɛla ba tan ni tan ye minnu bɛ baara kɛ ɲɔgɔn fɛ, k’a sɔrɔ u ma kɛ ni fɛnsɔrɔ musakaw ye minnu bɛ bɛn ɲɔgɔn ma. O kunba ye kunnafonidilanw sɔrɔli cogoyaw dilanni ye minnu bɛ cakɛda dancɛw bonya : ka kunnafonidilan funtenimaw to u fitininw na ani ka doncogo misaliw to u ka se ka fɔ ka ɲɛ, ka baara kɛ ni kalansenw ɲɔgɔnna ye walasa ka buffer pool doni tila, ani ka baarakɛminɛn-daɲɛw sɔrɔcogo dɔ sigi i n’a fɔ Redis kunnafonidilan ɲɛfɛ labanyɔrɔw kama minnu bɛ kunnafoni kelenw di baarakɛla caman ma waati kelen na.

Mewayz ye fɛn dilanni kɛ ni nin hakilila in tigitigi ye a hakili la . Ni jago modulu 207 jɛlen don minnu bɛ fanga di baarakɛla 138.000 ni kɔ ma, o yɔrɔ in ka kunnafonidilan dabɔra walasa kalan fanba ka baara kɛ ka bɔ cache la — ka jaabi waatiw to teliya la ani ka infrastructure musakaw to ka se ka fɔ ka ɲɛ n’i bɛ baara kɛ $19/kalo daminɛ bolodalen kan walima $49/kalo baarakɛlaw ka tier.

Ɲininkali minnu bɛ kɛ tuma caman na

ɲininkali marayɔrɔ balili bɛ kunnafonidilanw baarakɛcogo ɲɛ tuma bɛɛ wa ?

A tɛ kɛ tuma bɛɛ , nka sɛbɛnni-gɛlɛnw ta la a ka c' a la . Ɲininkali cakɛda bɛ diɲɛ mutex de wajibiya walasa ka bɛnkan sabati, o min bɛ kɛ buteli ye ɲɔgɔndanba kɔrɔ. MySQL 8.0 ye ɲininkali-yɔrɔ Bɔ pewu o sababu la. PostgreSQL ma ɲininkali-yɔrɔ jɔlen dɔ waleya abada, a y’a jigi da buffer pool ni application-layer caching kan o nɔ na. N’i ka kalan ni sɛbɛnni hakɛ ka ca ani n’i ka ɲininkaliw bɛ segin-ka-bɔ kosɛbɛ, ɲininkaliw marayɔrɔ bɛ se ka tɔnɔ lakikaw lase — n’o tɛ, i k’o tuning cɛsiri in bila buffer pool kɔnɔ.

ne bɛ se k' a dɔn cogo di ni n ka buffer pool hakɛ bɛ bɛn ?

I ka buffer pool hit ratio kɔlɔsi : ɲɛ ɲinini kɛmɛsarada la minnu bɛ baara kɛ ka bɔ pool la ni minnu bɛ disiki kalanni wajibiya. Ni hit ratio bɛ 95% duguma OLTP baarakɛta kan, o ye taamasyɛn ye walasa ka dɔ fara pool hakɛ kan. MySQL kɔnɔ, ɲininkali kɛ SHOW ENGINE INNODB STATUS ani ka buffer pool hit rate lajɛ. PostgreSQL kɔnɔ, pg_statio_user_tables filɛli bɛ heap blocks kalanlenw jira ka bɔ disiki la ni minnu bɛ baara kɛ ka bɔ buffer pool kɔnɔ. Aw ye aw jija k’aw ka baarakɛjɛkulu bɛɛ — a kana kɛ aw ka kunnafonidilan dafalen ye — ka to RAM kɔnɔ.

Jɛɲɔgɔnya jumɛn bɛ kasɛti layɛrɛw ni lokatɛri caman ka SaaS dannaya cɛ ?

Lokatɛri caman SaaS kɔnɔ, cakɛda layɛrɛw bɛ "sigiɲɔgɔn mankanba" gɛlɛyaw bali yɔrɔ min na lokatɛri kelen ka ɲininkali doni girinman bɛ baara kɛcogo tiɲɛ lokatɛri tɔw bɛɛ la. Locataire-aware application caching ni TTL-based invalidation bɛ locataire kelen-kelen bɛɛ ka kunnafoniw funteni mara Redis kɔnɔ, ka buffer pool degun dɔgɔya kosɛbɛ ka bɔ jatebɔsɛbɛnbaw la. Database-level connection pooling faralen ɲɔgɔn kan ni buffer pool sumalen ye, o b’a to burst baara min bɛ bɔ jatebɔsɛbɛn kelen o jatebɔsɛbɛn na, o tɛ ɲɛ jɛlenw bɔ cache kɔnɔ ani ka latency spikes kɛ plateforme bɛɛ kɔnɔ.

Cache layers tɛ database trivia ye — u ye architectural foundation ye min bɛ plateformew faranfasi minnu bɛ to teliya la scale la ni minnu bɛ infrastructure tasumafagalanw wajibiya tuma bɛɛ. N’i bɛ jago dɔ jɔ walima k’a ɲɛminɛ min mago bɛ yɔrɔ la min labɛnna kaban nin tiɲɛkow kama, Mewayz sɛgɛsɛgɛ app.mewayz.com — Module 207, plateforme kelen min bɛ bɛn ɲɔgɔn ma, min jɔra walasa ka baara kɛ ni dannaya ye k’a ta i ka baarakɛla fɔlɔ la ka se i ka ba kɛmɛnan ma.