Took the plunge during the maintenance window this month and compressed the top 5 tables in my buffer cache. Just compressing 5 tables I freed up 20-25% of my buffer cache (Oracle is managing my memory via AMM, and generally it allocates about 4gb to Shared Pool and 20 to the Buffer Cache)
| Uncompressed | Compressed | ||||
| Table | Block Count | % in Cache | Block Count | % in Cache | GB Saved |
| INV.MTL_MATERIAL_TRANSACTIONS | 200563 | 100 | 99263 | 100 | 0.77 |
| INV.MTL_TRANSACTION_ACCOUNTS | 210645 | 100 | 67470 | 100 | 1.09 |
| BOM.CST_ITEM_COST_DETAILS | 256731 | 99 | 50159 | 100 | 1.58 |
| INV.MTL_TRANSACTION_LOT_NUMBERS | 118761 | 100 | 48175 | 100 | 0.54 |
| WIP.WIP_TRANSACTIONS | 102027 | 100 | 42543 | 100 | 0.45 |
| Sum | 4.43 |
* Calculations assume the uncompressed object was 100% in cache, so may be off slightly, and are based on an 8k block size in case you want to check my math.
First impressions? Everything is running fine, if a little heavier on CPU and lighter on I/O. It’s really hard to quantify this early in the game. Next maintenance window, I’ll probably compress 50 more tables, and ultimately we’ll be really close to a “11i Database in Memory” environment.
