> - At this moment, there's another VM guest with 1 GB in size. > The database is 176 GB in size, used for datawarehouse. The > common accesses are transactional (short term requests) both > in queries and updates. > The results of our performance tests are bad (in words of our > DBA, because my problem is that I have no knowledge of > Oracle, so I only know what my DBA tells me). What I can > observe from Performance Toolkit is that we have a high rate > of pages moving from central to expanded, and from expanded > to central, causing a high CPU use.
How large is the virtual machine containing your database engine? Sounds like you're thrashing moving pages above/below the 2G line. Your DBA guy will probably think you're crazy, but one thing you might try is allocating several swap disks on VDISK (multiple small ones) for that guest, and reducing the size of the virtual machine to 256M or smaller. If you have difficulty, or you immediately go heavily into swapping, increase the size of the virtual machine by 32M increments until you see about 10-15% swap usage. You will also want to _reduce_ the amount of internal buffering that Oracle does to minimize virtual storage usage inside the DB machine. The object is to get the virtual machine size working set to a manageable level for VM, which will help reduce the amount of thrashing going on between main and xstore. Much of the Oracle tuning "wisdom" is geared toward memory-rich, I/O poor environments. This is exactly the opposite situation -- memory is shared (and thus an expensive resource), and you have very fast I/O. ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
