Tare da wannan ɗan gajeren rubutu zan so in kawar da rashin fahimta guda ɗaya da ke da alaƙa da nazarin bayanan AWR da ke gudana akan Oracle Exadata. Kusan shekaru 10, koyaushe ina fuskantar tambaya: menene gudummawar Software Exadata ga yawan aiki? Ko yin amfani da sabbin kalmomin da aka ƙirƙira: yaya “gwani” aikin wani takamaiman bayanai ne?
Sau da yawa wannan madaidaicin tambayar, a ganina, ana amsawa ba daidai ba dangane da kididdigar AWR. Yana gabatar da hanyar jira na tsarin, wanda ke ɗaukar lokacin amsawa azaman jimlar lokacin aiki na masu sarrafawa (DB CPUs) da lokacin jira na azuzuwan daban-daban.
Tare da zuwan Exadata, takamaiman tsammanin tsarin da ke da alaƙa da aikin Exadata Software ya bayyana a kididdigar AWR. A matsayinka na mai mulki, sunayen irin waɗannan jirage suna farawa da kalmar "cell" (ana kiran uwar garken Storage Exadata a cell), wanda mafi yawansu suna jira tare da sunaye masu bayyana kansu "cell smart table scan", "cell multiblock". karantawa ta zahiri” da “karantawa tantanin halitta guda ɗaya”.
A mafi yawan lokuta, rabon irin wannan Exadata yana jira a cikin jimlar lokacin amsawa kaɗan ne, sabili da haka ba sa ma faɗa cikin Top10 Foreground Events by Total Wait Time sashe (a wannan yanayin, kuna buƙatar nemo su a cikin Jiran Farko). Sashen abubuwan). Tare da wahala mai girma, mun sami misalin AWR na yau da kullun daga abokan cinikinmu, wanda a ciki aka haɗa tsammanin Exadata a cikin sashin Top10 kuma a cikin duka ya kai kusan 5%:
Event
Jira
Jimlar Lokacin Jira (minti)
Matsakaicin Jira
% DB lokaci
Jira Class
DB CPU
115.2K
70.4
SQL * Net ƙarin bayanai daga dblink
670,196
5471.5
8.16ms
3.3
Network
tantanin halitta tubalan jiki karatu
5,661,452
3827.6
676.07us
2.3
Mai amfani I/O
Daidaita daidaitawar ASM
4,350,012
3481.3
800.30us
2.1
Other
cell multiblock karatun jiki
759,885
2252
2.96ms
1.4
Mai amfani I/O
hanyar kai tsaye karanta
374,368
1811.3
4.84ms
1.1
Mai amfani I/O
SQL * Saƙon Net daga dblink
7,983
1725
216.08ms
1.1
Network
cell smart table scan
1,007,520
1260.7
1.25ms
0.8
Mai amfani I/O
hanyar kai tsaye karanta temp
520,211
808.4
1.55ms
0.5
Mai amfani I/O
enq: TM - jayayya
652
795.8
1220.55ms
0.5
Aikace-aikace
Ana samun sakamako masu zuwa daga irin waɗannan ƙididdiga na AWR:
1. Gudunmawar sihirin Exadata ga aikin bayanai ba ta da yawa - bai wuce 5% ba, kuma ma'ajin yana “ɓata” da kyau.
2. Idan irin wannan bayanan an canza shi daga Exadata zuwa tsarin gine-ginen "server + array", to aikin ba zai canza da yawa ba. Domin ko da wannan tsararriyar ta juya ta zama sau uku a hankali fiye da tsarin ajiya na Exadata (wanda ba shi yiwuwa ga zamani All Flash arrays), sannan ninka 5% da uku za mu sami karuwa a cikin rabon I/O yana jira zuwa 15% - database zai tsira daga wannan!
Duk waɗannan abubuwan biyu ba daidai ba ne, haka kuma, suna karkatar da fahimtar ra'ayin da ke tattare da software na Exadata. Exadata ba kawai yana samar da I/O mai sauri ba, yana aiki ne daban-daban idan aka kwatanta da na al'ada uwar garken + tsarin gine-gine. Idan aikin bayanan da gaske ya kasance "wanda aka cire", to ana canza ma'anar SQL zuwa tsarin ajiya. Sabbin ma'ajiya, godiya ga wasu hanyoyi na musamman (musamman Exadata Storage Indexes, amma ba kawai), sami mahimman bayanai da kansu kuma aika DB zuwa sabobin. Suna yin wannan da kyau sosai, don haka rabon Exadata na yau da kullun yana jira a cikin jimlar lokacin amsawa kaɗan ne.
Ta yaya wannan rabon zai canza a wajen Exadata? Ta yaya hakan zai shafi aikin rumbun adana bayanai gaba daya? Gwaji zai fi amsa waɗannan tambayoyin. Misali, jiran “cancantar tebur mai wayo” a wajen Exadata na iya juya zuwa irin wannan Cikakken Cikakken Scan na Tebu wanda I/O ke ɗaukar duk lokacin amsawa kuma aikin yana raguwa sosai. Shi ya sa ba daidai ba ne, lokacin da ake nazarin AWR, a yi la'akari da jimillar kashi na tsammanin Exadata a matsayin gudunmawar sihirinsa ga aiki, har ma fiye da haka don amfani da wannan kashi don hasashen aiki a wajen Exadata. Don fahimtar yadda "daidai" aikin bayanan, kuna buƙatar nazarin ƙididdigar AWR na sashin "Ayyukan Ayyukan Misali" (akwai ƙididdiga masu yawa tare da sunayen bayyana kansu) kuma kwatanta su da juna.
Kuma don fahimtar yadda bayanan da ke waje da Exadata zai ji, yana da kyau a yi clone na bayanai daga ajiyar ajiya akan gine-ginen da aka yi niyya da kuma nazarin aikin wannan clone a ƙarƙashin kaya. Masu Exadata, a matsayin mai mulkin, suna da wannan damar.
Author: Alexey Struchenko, shugaban sashen bayanai na Jet Infosystems
source: www.habr.com