Requests for I/O operations on a logical volume can be biased toward read operations, write operations, or a balanced mix of the two types of operation. You can use the bias of the type of I/O operation (in conjunction with the access bias property) to help optimize the performance of a logical volume.
As a Quality of Service (QoS) property, the I/O bias indicates to the system the type of I/O operation that is the most common or is the most expected for a particular volume:
- Read
Indicates that most of the access requests are for read operations.
- Write
Indicates that most of the access requests are for write operations.
- Mixed
Indicates that the number of access requests are similar for read operations and for write operations.
Important! If you choose Random as the access bias and Write as the I/O Bias, and if you choose single redundancy for data protection, the Oracle FS System creates the logical volume using RAID 10 (mirrored data). This RAID level enhances the write performance but uses twice the capacity.
For write requests, the system stores all of the user data and all of the system metadata in mirrored copies of the journal. One copy is maintained in the regular cache on the owning
Controller. The mirror copy is maintained in one of the following locations:
The memory of the partner Controller that is backed by an energy storage module (ESM). All host write operations are copied to the ESM-backed, non-volatile cache in the partner Controller. This type of operation is often refered to as a write back cache.
The virtual LUN (VLUN) that is reserved on the physical storage for the logical volume, if the partner Controller is unavailable for the write operation. Write operations to this copy are equivalent to write-through cache.
Write operations from the journal to permanent physical storage are equivalent to write-through cache. The system flushes the user data and the corresponding metadata as a unit to the physical storage.