Home | Contents | Help   Raidhelp     Controller Redundancy: Active-Passive Architecture An active-passive dual controller scheme is one in which a master or Primary controller services all host I/O requests to a pool of disk drives whilst the slave or Secondary controller remains on stand-by and does not assist in servicing I/O requests from the host or network. The controllers may control a single or multiple number of logical volumes. Each logical volume may also reside on the same array or on a separate independent array. The logical volume configuration across all attached drives is known by both the master controller and the slave controller as they are both physically connected to the same arrays and have access to the same logical volumes. Both controllers are physically connected to the same I/O bus at all times. The slave controller constantly monitors the availability of the master controller. In the event the master controller fails, the slave controller transparently takes over servicing host I/O requests to any logical drives residing on the pool of disk drives. This is accomplished by taking over the role of the master controller and accepting all I/O requests from all hosts that are accessing the array.  There are two main methods in which Primary and Secondary controllers can be configured depending on manufacturer: Impersonation and Single Virtual Device.  Impersonation Some Secondary controllers are invisible to the hosts until a Primary controller failure where they then impersonate the Primary controller and respond to I/O requests addressed to the Primary controller. The Secondary controller becomes the active controller and the Primary controller is taken off-line or relegated to the position of the passive controller. Single Virtual Device The two controllers are linked via a dedicated port and configure themselves as a single virtual device to the host systems at the initial installation before any failure has even occurred. No impersonation is necessary as both controllers are seen by the hosts as the same device anyway. Again, the Secondary controller becomes the active controller and the Primary controller is taken off-line or relegated to the position of the passive controller. Using operating system-independent controllers means that host and client systems should never be aware that they are using a RAID system, therefore any hosts or clients should also be completely unaware that any error has occurred and simply continue to access the data through normal I/O requests. In the case of a SCSI based controller for example, the host simply sends I/O requests to a specific LUN and SCSI ID target. The active controller simply deals with all I/O requests to that target address.   The fact that the surviving controller is impersonating the failed controller indicates that the host does not have to do anything special to guarantee data availability to user processes. Related topics: Controller Redundancy Active-Active Architectures Four States of an Array   Raidhelp   RAIDhelp© Copyright 1999-2004 Antony Kershaw