IBM F80 System Power Control Network Spcn, Disk Redundancy Mirroring, RAID, Dual Controllers

handling mode that allows the bridge chip to freeze access to an adapter when a PCI bus error occurs on the interface between that adapter and bridge chip. In this frozen mode, DMAs7 are blocked, stores to that device address space are discarded, and loads result in a return value of all 1s. Device drivers can be programmed to look for these dummy responses on loads and can attempt recovery. The AIX support for this function is not available yet.

System Power Control Network (SPCN)

SPCN consists of a set of power/environmental controllers, interconnected by a set of serial communication links. In Model F80 systems, the SPCN function is integrated into the service processor and provides the following functions:

•Powering all the system parts up or down, when requested.

The SPCN hardware has connections to the VPD that is resident on each of

the pluggable cards and the backplane. The VPD is located on each of the cards in the form of an I2C chip. This chip is accessed during initial power on

sequence and the data contents are read by the service processor. Using this function, the service processor decides not to use components that are marked bad.

•Powering down all the system parts on critical power faults

•Monitors power, fans, and thermal conditions in the system for problem conditions, which result in an EPOW. EPOW stands for environmental and power off warnings and is a function to inform the service processor or the operating system early, about an event that happened in the hardware. There are different warnings; such as cooling warnings or power fail warnings which result in entries in the error log. If there is a serious error, such as the temperature reaches a specific limit, the system will be shutdown.

•Reporting power and environmental faults, as well as faults in the SPCN network itself, on operator panels and through the service processor

•Assigning and writing location information into various VPD elements in the system.

Disk Redundancy (Mirroring, RAID, Dual Controllers)

RS/6000 and AIX provide a number of options for increasing the robustness of storage subsystems, all of which involve some level of redundancy of disks and/or adapters.

AIX disk mirroring provides the ability to define transparent double or triple redundancy of disk data by mapping disk write data to two or three physical disks. On disk reads, the request is issued to all disks in the mirror group, and the first error-free response is returned, which also has some performance benefits. If one of the disks fails, the data is still readable from the other disk(s).

There are also customer options for SCSI and SSA RAID controller adapters, which can provide the same protection with better performance and less redundancy overhead. Also available are storage subsystems that provide under-the-covers redundancy for high availability.

To provide protection against adapter failures, AIX also supports dual-controller options where the same disk subsystem can be accessed through both a primary adapter path and through a backup adapter path if the primary fails.

7Direct memory access (DMA)

IBM RS/6000 7025 Model F80 Server 15