SCSI Domain

A SCSI Domain, in simple terms, is a grouping of SCSI devices that can see each other and are allowed to communicate with each other.

If you were to look into the actual SCSI Architectural Model (SAM) document (see www.t10.org), SAM says that a SCSI Domain consists a system that has at least one SCSI Initiator device and one SCSI Target device that are connected through a service delivery subsystem.  In the good old days of parallel SCSI, where devices were identified by their SCSI ID, and had to be directly attached to a SCSI cable (or backplane), that would mean that a maximum of 16 devices could be in a single domain.  That was simple compared to today’s SAN environments.

But with the advent of other storage technologies that use this same model (like Fibre Channel, iSCSI, and SAS), but use different physical signaling methods (serial and switched vs. parallel), and where devices and ports are identified based on their world wide name (WWN) or world wide identifier (WWID), then the number of devices in a single domain can grow to an alarming size.

In an attempt to keep these SCSI domains under control, the concept of “zoning” devices into groups came about.  Most of the zoning is done within components of the infrastructure, such as SAN switches (for Fibre Channel) or expanders (for SAS), but the concept is the same: being able to keep the SCSI domain to a manageable size and control what host systems (initiators) can see particular storage devices (targets).

In some cases, the storage devices themselves can also manage what host computer systems can “see” different parts of the storage sub-systems.  This is mainly done on enterprise-class storage systems (think big RAID systems) using specialized storage management software.  Again, it is all about limiting visibility and controlling access based on certain parameters.