![\"AWR:](\"\/wp-content\/uploads\/2020\/09\/b0ca997a4bd415723ae56c183e2dd42f.jpeg\")
<\/p>\nMe k\u00ebt\u00eb post t\u00eb vog\u00ebl doja t\u00eb shpjegoja nj\u00eb keqkuptim n\u00eb lidhje me analiz\u00ebn e databazave AWR q\u00eb funksionojn\u00eb n\u00eb Oracle Exadata. Gati 10 vjet kam hasur vazhdimisht pyetjen: cila \u00ebsht\u00eb kontributi i Exadata Software n\u00eb performanc\u00ebn? Ose me fjal\u00eb t\u00eb reja: sa \u2018exadata\u2019 \u00ebsht\u00eb puna e nj\u00eb databaze t\u00eb caktuar?<\/p>\n
<\/p>\n
Shpesh, p\u00ebr k\u00ebt\u00eb pyetje t\u00eb sakt\u00eb, sipas mendimit tim, jepet nj\u00eb p\u00ebrgjigje e gabuar e cila referohet n\u00eb statistikat AWR. Ajo paraqet nj\u00eb metod\u00eb t\u00eb pritjeve sistematike, q\u00eb e interpreton koh\u00ebn e p\u00ebrgjigjes si shum\u00ebn e koh\u00ebs s\u00eb pun\u00ebs s\u00eb procesor\u00ebve (DB CPU) dhe koh\u00ebs s\u00eb pritjeve nga klasat e ndryshme.<\/p>\n
Me shfaqjen e Exadata n\u00eb statistikat AWR u shfaq\u00ebn pritje sistematike specifike, t\u00eb lidhura me pun\u00ebn e Exadata Software. N\u00eb p\u00ebrgjith\u00ebsi, emrat e k\u00ebtyre pritjeve fillojn\u00eb me fjal\u00ebn 'cell' (servery Exadata Storage quhet cell), nd\u00ebr to, pritjet m\u00eb t\u00eb zakonshme jan\u00eb me emra t\u00eb kuptuesh\u00ebm si 'cell smart table scan', 'cell multiblock physical read' dhe 'cell single block physical read'.<\/p>\n
N\u00eb shumic\u00ebn e rasteve, pjesa e k\u00ebtyre pritjeve Exadata n\u00eb koh\u00ebn totale t\u00eb p\u00ebrgjigjes \u00ebsht\u00eb e vog\u00ebl, dhe prandaj ato madje nuk arrijn\u00eb t\u00eb p\u00ebrfshihen n\u00eb seksionin Top10 ngjarjet e p\u00ebrparme sipas koh\u00ebs totale t\u00eb pritjes (n\u00eb k\u00ebt\u00eb rast, ato duhet t\u00eb k\u00ebrkohen n\u00eb seksionin ngjarjet e pritjes s\u00eb p\u00ebrparme). Ne me v\u00ebshtir\u00ebsi gjet\u00ebm nj\u00eb shembull t\u00eb AWR-\u00ebs ditor nga klient\u00ebt tan\u00eb, ku pritjet Exadata arrit\u00ebn n\u00eb seksionin Top10 dhe p\u00ebrb\u00ebnin rreth 5% t\u00eb gjithsej:<\/p>\n
Ngjarja<\/strong><\/p>\n Pritjet<\/strong><\/p>\n Koha totale e pritjes (sekond\u00eb)<\/strong><\/p>\n Pritja mesatare<\/strong><\/p>\n % Koha DB<\/strong><\/p>\n Klasa e pritjes<\/strong><\/p>\n DB CPU<\/p>\n <\/p>\n 115.2K<\/p>\n <\/p>\n 70.4<\/p>\n <\/p>\n SQL*Net m\u00eb shum\u00eb t\u00eb dh\u00ebna nga dblink<\/p>\n 670,196<\/p>\n 5471.5<\/p>\n 8.16ms<\/p>\n 3.3<\/p>\n Network\nglobal\n log \/dev\/log local0\n log \/dev\/log local1 notice\n chroot \/var\/lib\/haproxy\n stats timeout 30s\n user haproxy\n group haproxy\n daemon\n\ndefaults\n log global\n mode http\n option httplog\n option dontlognull\n timeout connect 5000\n timeout client 50000\n timeout server 50000\n\nfrontend http_front\n bind *:80\n stats uri \/haproxy?stats\n default_backend http_back\n\nbackend http_back\n balance roundrobin\n server server_name1 private_ip1:80 check\n server server_name2 private_ip2:80 check<\/p>\n cell single block physical read<\/p>\n 5,661,452<\/p>\n 3827.6<\/p>\n 676.07us<\/p>\n 2.3<\/p>\n P\u00ebrdorues I\/O<\/p>\n Sync ASM rebalance<\/p>\n 4,350,012<\/p>\n 3481.3<\/p>\n 800.30us<\/p>\n 2.1<\/p>\n Tjet\u00ebr<\/p>\n cell multiblock physical read<\/p>\n 759,885<\/p>\n 2252<\/p>\n 2.96ms<\/p>\n 1.4<\/p>\n P\u00ebrdorues I\/O<\/p>\n rr\u00ebfimi direct path read<\/p>\n 374,368<\/p>\n 1811.3<\/p>\n 4.84ms<\/p>\n 1.1<\/p>\n P\u00ebrdorues I\/O<\/p>\n SQL*Net mesazhi nga dblink<\/p>\n 7,983<\/p>\n 1725<\/p>\n 216.08ms<\/p>\n 1.1<\/p>\n Network\nglobal\n log \/dev\/log local0\n log \/dev\/log local1 notice\n chroot \/var\/lib\/haproxy\n stats timeout 30s\n user haproxy\n group haproxy\n daemon\n\ndefaults\n log global\n mode http\n option httplog\n option dontlognull\n timeout connect 5000\n timeout client 50000\n timeout server 50000\n\nfrontend http_front\n bind *:80\n stats uri \/haproxy?stats\n default_backend http_back\n\nbackend http_back\n balance roundrobin\n server server_name1 private_ip1:80 check\n server server_name2 private_ip2:80 check<\/p>\n cell smart table scan<\/p>\n 1,007,520<\/p>\n 1260.7<\/p>\n 1.25ms<\/p>\n 0.8<\/p>\n P\u00ebrdorues I\/O<\/p>\n rr\u00ebfimi direct path read temp<\/p>\n 520,211<\/p>\n 808.4<\/p>\n 1.55ms<\/p>\n 0.5<\/p>\n P\u00ebrdorues I\/O<\/p>\n enq: TM \u2014 kontestimi<\/p>\n 652<\/p>\n 795.8<\/p>\n 1220.55ms<\/p>\n 0.5<\/p>\n Aplikacioni<\/p>\n Nga nj\u00eb statistik\u00eb e till\u00eb AWR shpesh b\u00ebhen p\u00ebrfundime t\u00eb tilla:<\/p>\n 1. Kontributi i magjis\u00eb Exadata n\u00eb performanc\u00ebn e baz\u00ebs s\u00eb t\u00eb dh\u00ebnave nuk \u00ebsht\u00eb i lart\u00eb \u2014 nuk kalon 5%, dhe baza e t\u00eb dh\u00ebnave 'ekzadatohet' keq.<\/p>\n 2. N\u00ebse nj\u00eb baz\u00eb e till\u00eb transferohet nga Exadata n\u00eb arkitektur\u00ebn klasike 'server + sistem t\u00eb ruajtjes', performanca nuk do t\u00eb ndryshoj\u00eb ndjesh\u00ebm. Sepse, edhe n\u00ebse ky sistem i ruajtjes do t\u00eb ishte tri her\u00eb m\u00eb i ngadalsh\u00ebm se sistemi i ruajtjes Exadata (\u00e7far\u00eb mund t\u00eb duket e pabesueshme p\u00ebr sistemet p\u00ebr ruajtjen All Flash), duke shum\u00ebzuar 5% me tre ne do t\u00eb arrijm\u00eb nj\u00eb rritje t\u00eb pjes\u00ebs s\u00eb pritjeve t\u00eb input-output deri n\u00eb 15% \u2014 t\u00eb till\u00eb nj\u00eb baz\u00eb t\u00eb dh\u00ebnash me siguri do ta p\u00ebrballonte!<\/p>\n T\u00eb dy k\u00ebto dalje jan\u00eb t\u00eb pasakta, madje ato shtremb\u00ebrojn\u00eb kuptimin e ides\u00eb q\u00eb q\u00ebndron pas Exadata Software. Exadata nuk siguron thjesht hyrje-dalje t\u00eb shpejt\u00eb, ajo funksionon n\u00eb m\u00ebnyr\u00eb t\u00eb ndryshme krahasuar me arkitektur\u00ebn klasike \"server + grumbull\". N\u00ebse puna e baz\u00ebs s\u00eb t\u00eb dh\u00ebnave v\u00ebrtet \"ekzadatiset\" \u2014 at\u00ebher\u00eb logjika SQL transferohet n\u00eb sistemin e ruajtjes. server\u00ebt<\/a> Fal\u00eb nj\u00eb s\u00ebr\u00eb mekanizmash special\u00eb (n\u00eb radh\u00eb t\u00eb par\u00eb Exadata Storage Indexes, por jo vet\u00ebm) ata vet\u00eb gjejn\u00eb t\u00eb dh\u00ebnat e nevojshme dhe i d\u00ebrgojn\u00eb n\u00eb serverat DB. Ata e b\u00ebjn\u00eb k\u00ebt\u00eb mjaft efektivisht, ndaj pjesa e pritjeve karakteristike Exadata n\u00eb koh\u00ebn totale t\u00eb p\u00ebrgjigjes \u00ebsht\u00eb e vog\u00ebl. <\/p>\n Si do t\u00eb ndryshoj\u00eb kjo pjes\u00eb jasht\u00eb Exadata? Si do t\u00eb ndikoj\u00eb kjo n\u00eb performanc\u00ebn e baz\u00ebs s\u00eb t\u00eb dh\u00ebnave n\u00eb p\u00ebrgjith\u00ebsi? M\u00eb mir\u00eb se \u00e7dokush mund ta p\u00ebrgjigjet testimi. P\u00ebr shembull, pritja \"cell smart table scan\" jasht\u00eb Exadata mund t\u00eb shnd\u00ebrrohet n\u00eb nj\u00eb skan t\u00eb r\u00ebnd\u00eb Table Full Scan, aq sa hyrje-dalje do t\u00eb marr\u00eb t\u00eb gjith\u00eb koh\u00ebn e p\u00ebrgjigjes dhe performanca do t\u00eb p\u00ebrkeq\u00ebsohet dramatisht. Prandaj, \u00ebsht\u00eb e gabuar n\u00eb analiz\u00ebn AWR t\u00eb konsiderohet p\u00ebrqindja totale e pritjeve Exadata si nj\u00eb kontribut t\u00eb magjis\u00eb s\u00eb saj n\u00eb performanc\u00eb dhe aq m\u00eb tep\u00ebr t\u00eb p\u00ebrdoret kjo p\u00ebrgjigje p\u00ebr t\u00eb parashikuar performanc\u00ebn jasht\u00eb Exadata. P\u00ebr t\u00eb kuptuar sa \"ekzadatiset\" puna e baz\u00ebs duhet t\u00eb studiohen statistikat AWR n\u00eb seksionin \u201cInstance Activity Stats\u201d (aty ka shum\u00eb statistika me emra tregues) dhe ato duhet t\u00eb krahasohen me nj\u00ebra-tjetr\u00ebn.<\/p>\n Dhe p\u00ebr t\u00eb kuptuar si do t\u00eb ndjehet baza e t\u00eb dh\u00ebnave jasht\u00eb Exadata, \u00ebsht\u00eb m\u00eb mir\u00eb t\u00eb b\u00ebni nj\u00eb klon t\u00eb baz\u00ebs nga backup n\u00eb arkitektur\u00ebn e synuar dhe t\u00eb analizoni performanc\u00ebn e k\u00ebtij kloni n\u00ebn ngarkes\u00eb. Nj\u00eb mund\u00ebsi e till\u00eb zakonisht \u00ebsht\u00eb n\u00eb dispozicion p\u00ebr ata q\u00eb kan\u00eb Exadata.<\/p>\n