Users Guide
• No Read Ahead — Disables the Read-Ahead capability.
• Adaptive Read Ahead — Read Ahead Mode is used when the two most recent disk accesses occurred in sequential sectors,
otherwise the controller reverts to No Read Ahead mode.
NOTE: The default read cache setting for virtual disks is Adaptive Read Ahead.
Physical disk write cache policy
Management utilities provide the option to modify disk cache policies for physical disks associated with a virtual disk. The following options
are available for the Shared PERC controller:
Option Description
Enabled Enables physical disk cache for drives associated with the selected virtual disk.
Disabled Disables physical disk cache for drives associated with the selected virtual disk.
Default The default cache policy on the physical disk is used. See the physical disk documentation for details on the default
setting.
WARNING: Enabling physical disk cache is not recommended for disks with critical data as a power failure may result in the loss
of data prior to it being ushed to the persistent media.
NOTE: Any change to the physical disk cache applies to all virtual disks in an entire disk group. It is not possible to have dierent
cache policies on virtual disks from the same disk group.
Fault tolerance
The fault tolerance features of Shared PERC 8 are as follows:
• Support for Self Monitoring and Reporting Technology (SMART)
• Support for Patrol Read
• Physical disk failure detection
• Physical disk rebuild using hot spares
• Controller Cache Preservation
• Battery and non-volatile cache backup of controller cache to protect data
• Detection of batteries with low charge after boot up
• Support for controller failover in systems with Fault Tolerant Shared PERC 8 card conguration.
The following sections describe some methods to achieve fault tolerance.
The SMART feature
The SMART feature monitors certain physical aspects of all motors, heads, and physical disk electronics to help detect predictable physical
disk failures. Data on SMART-compliant physical disks can be monitored to identify changes in values and determine whether the values are
within threshold limits. Many mechanical and electrical failures display some degradation in performance before failure.
A SMART failure is also referred to as predicted failure. There are numerous factors that are predicted physical disk failures, such as a
bearing failure, a broken read/write head, and changes in spin-up rate. In addition, there are factors related to read/write surface failure,
such as seek error rate and excessive bad sectors.
NOTE
: For detailed information on SCSI interface specications, see t10.org and for detailed information on SATA interface
specications, see t13.org.
Shared PERC 8 card features 17