The Oracle Database has a layered architecture that includes the Oracle Disk Manager (ODM). The ODM provides a file management module that lets the Oracle Database use a local file system, a raw disk partition, or NFS server to store database information.
To increase database performance, the ODM interface lets the Oracle Database pass information along with each I/O request. This information defines several attributes associated with the I/O such as the file type associated with the I/O request. This lets data file and database log file writes be handled differently.
The new OISP allows the Oracle Database NFSv4 client to pass ODM optimization information to the NFSv4 server of the ZFS Storage Appliance. The ZFS Storage Appliance takes advantage of the ODM optimization information to simplify database configuration and to further increase database performance.
There are two Oracle Intelligent Storage Protocol features:
Automatically setting the Optimal file record size for new database files
Automatically using the optimal write bias (ZFS Latency or Throughput) for each write request
The Oracle dNFS client passes the optimal record size to the ZFS Storage Appliance for each NFSv4 write request. The ZFS Storage Appliance NFSv4 server passes the record size to the ZFS file system with the I/O request. The ZFS file system then bypasses the default file system record size and uses the record size value passed with the I/O request. The record size can only be set for newly created files. If a file already exists the record size will not be changed.
The Oracle dNFS client passes the optimal write bias to the ZFS Storage Appliance for each NFSv4 write request. The ZFS Storage Appliance NFSv4 server passes the write bias to the ZFS file system with the I/O request. The ZFS file system then bypasses the default file system write bias and attempts to use the write bias value passed with the I/O request. Depending on the state of the ZFS file system the write bias sent with the I/O request may be ignored.