Just moved some databases from Solaris 10 on Sparc servers to Solaris 11 on new Sparc servers. Some were lateral moves, some were 11.2.x to 12.1.0.2 upgrades.
Previously all databases resided on the one box, new server is sliced up into 3 VM's, each with processors & memory allocated so the databases can run with the same parameters as before. Also are using the ZFS filesystem on the new servers, which from what I understand is optimized to work with Oracle database.
One thing very noticeable across the board is slower performance, both in queries and even in database startup sequence. For example, one database that takes 22 seconds to start on stand alone server now takes 223 seconds to start on new server. The startup sequence is noticeably slower for all databases.
When comparing side by side queries the new server is initially a bit faster than the old, but subsequent runs of the same queries the old server improve drastically (as expected) whereas the new server the subsequent runs are only marginally faster than the first, even though buffer cache ratio is 100%.
I have a suspicion the OVM's are not managing memory properly - has anyone else run into this, and if so -how was it addressed?
thanks in advance..