Krishna Manoharan wrote,on my timestamp of 7/10/2008 10:51 AM:
1. Huge pages are used for SGA only. Don't reserve more than you intend to allocate to SGA (SGA_MAX_SIZE). Otherwise, you are going to see heavy swapping.
That would be highly dependent on the implementation.
2. Huge pages cannot be paged out. They are locked in memory (similar to ISM on Solaris) and even if un-used, not freed back to the OS. No other application other than Oracle can use this memory.
If they cannot be paged out, how can they result in heavy swapping? Like I said: dependent on implementation. Solaris is just one of many.
5. Huge pages are more efficient because of the bigger pagesize (2M) and also help reduce TLB misses (assuming you are having a system with more than 16GB Memory).
Actually, I've seen them be useful at much less than 16GB. Typical TLBs can cope well with 32-bit size memories, or in other words: 4GB. After that, they start to show locate speed issues: simply, way too many page entries. The hugepage size reduces the number of required entries in the TLB for a given memory size. With Red Hat, hugepages show CPU usage improvement after 4GB - assuming of course one has an SGA of matching size and a workload that can show problems. -- Cheers Nuno Souto in sunny Sydney, Australia dbvision@xxxxxxxxxxxx -- //www.freelists.org/webpage/oracle-l