🧠⚙️ 8) Mémoire & I/O (SGA / PGA / Storage)
Objectif : dimensionner et stabiliser la perf système. Ici on relie : SGA (buffer/cache/shared pool), PGA (work areas), TEMP/UNDO, redo log contention, et stockage/ASM (latence disque).
8.1
Sizing : méthode (preuve → ajustement)
Ce qu’on mesure avant de toucher la mémoire : DB Time, waits, read profile, workload.
SizingEvidenceSafe
8.2
SGA : buffer cache & shared pool
Logical/physical reads, cache “utile”, parsing, cursor sharing, invalidations.
SGACacheParsing
8.3
PGA & work areas : sorts / hash joins
Spill to TEMP, workarea sizing, auto PGA, risques d’overcommit.
PGAWorkareaTEMP
8.4
TEMP & UNDO : pression, erreurs, tuning
Temp usage, undo retention, ORA-01652/01555, batch & long queries.
TEMPUNDOStability
8.5
Redo log contention : LGWR & commit path
log file sync, log buffer space, parallel write, commits trop fréquents.
RedoCommitLGWR
8.6
Storage/ASM : latence & distribution I/O
Latence read/write, diskgroups, rebalance, hotspots, preuve vs “feeling”.
ASMLatencyHot disks
A
Diagrammes (Mermaid-ready)
SGA/PGA → TEMP → I/O, et boucle de sizing “safe”.
copiablepédago
B
Commandes & vues (memory/I/O)
V$SGA*, V$PGASTAT, V$MEMORY_*, V$TEMPSEG_USAGE, V$UNDOSTAT, ASM, I/O waits.
V$AWRRunbook
