G
George Potemkin
Guest
You can remove any blocks in buffer pool out of the LRU chain. Each buffer header has a field called "Skips" - it's a "countdown" for the block before the block's access will again change LRU chain. We can set the "countdown" equal to its maximum value (two billions minus one). In most cases it will mean "never change LRU chain". For a short period of time set the -lruskips to 2147483647, read the blocks and reset the -lruskips back to zero. Check promon/R&D/debghb/6/1. Status: Cache Entries or promon/R&D/debghb/6/4. Status: Lru Chains Num DBKEY Area Hash T S Usect Flags Updctr Lsn Chkpnt Lru Skips 33 64 1 139 O 0 L 4 0 0 0 2147483647 34 64 6 824 O 0 L 1385 0 0 0 2147483647 35 64 7 74 O 0 L 43 0 0 0 2147483647 36 64 8 211 O 0 L 10 0 0 0 2147483647 37 2 9 348 O 0 L 6 0 0 0 2147483647 38 2 10 485 O 0 L 6 0 0 0 2147483647 39 64 11 622 O 0 L 6 0 0 0 2147483647 Last column marks the blocks "who want live forever" off the LRU chain. LruShot.p does this trick for a few special blocks (including the ACO blocks): ftp.progress-tech.ru/.../LruShot.p For example, it would be good idea to put all metachema data in such "alternate buffer pool" because all sessions read a lot of blocks during db connection and sometime the connections should be fast as possible. It's easy to create the dynamically allocated alternate buffer pools right after db startup.
Continue reading...
Continue reading...