Active/passive means that one controller is active to process I/O requests, while the other is idle in standby mode ready to take over I/O activity should the active primary controller fail or be taken offline. An active/active configuration enables both controller nodes to process I/Os, and provide a standby capability for the other.
Generally, an active/active storage system involves a battery backed mirrored cache, in which the contents of a controller's cache is mirrored to another for data integrity and availability. For example, if an active/active controller has a total of 4 gigabytes (GB) of cache (e.g., 2 GB per controller), the usable amount of cache would be 50% or 2 GB due to mirroring.
Typically, a LUN will be accessed primarily via a specific controller to maintain cache coherency in what is known as LUN or cache affinity. Some controllers enable either controller to respond to a server's request for an I/O for a LUN. However, the actual I/O is resolved by a designated controller in what is known as I/O shipping. As to which approach is better, if you can afford to have one controller idle, an active/passive configuration gives predictable performance during a failover. Many NAS vendors use this model, which is often referred to as a NAS cluster. An active/active configuration enables both controllers to be active to boost performance during normal operation. However, during a failover scenario depending on the combined workload, performance could degrade.
Do you know…
This was first published in July 2006