Technical data

13-4 Fabric OS Administrator’s Guide
Publication Number: 53-0000518-09
Zoning Concepts
13
Table 13-2 Approaches to Fabric-Based Zoning
Zoning by:
Single HBA Zoning by single HBA most closely re-creates the original SCSI bus. Each zone
created has only one HBA (initiator) in the zone; each of the target devices is
added to the zone. Typically, a zone is created for the HBA and the disk storage
ports are added. If the HBA also accesses tape devices, a second zone is created
with the HBA and associated tape devices in it. In the case of clustered systems, it
could be appropriate to have an HBA from each of the cluster members included
in the zone; this is equivalent to having a shared SCSI bus between the cluster
members and presumes that the clustering software can manage access to the
shared devices. In a large fabric, zoning by single HBA requires the creation of
possibly hundreds of zones; however, each zone contains only a few members.
Zone changes affect the smallest possible number of devices, minimizing the
impact of an incorrect zone change. This zoning philosophy is the preferred
method.
Application Zoning by application typically requires zoning multiple, perhaps incompatible,
operating systems into the same zones. This method of zoning creates the
possibility that a minor server in the application suite could disrupt a major server
(such as a Web server) disrupting a data warehouse server. Zoning by application
can also result in a zone with a large number of members, providing greater
susceptibility to administrative errors, such as registered state change
notifications (RSCNs) going out to a larger group than necessary.
Operating system Zoning by operating system has issues similar to zoning by application. In a large
site, this type of zone can become very large and complex. When zone changes
are made, they typically involve applications rather than a particular server type.
If members of different operating system clusters can see storage assigned to
another cluster, they might attempt to own the other cluster’s storage and
compromise the stability of the clusters.
Port allocation Avoid zoning by port allocation unless the administration team has very rigidly
enforced processes for port and device allocation in the fabric. It does, however,
provide some positive features. For instance, when a storage port, server HBA, or
tape drive is replaced, the change of WWN for the new device is of no
consequence. As long as the new device is connected to the original port, it
continues to have the same access rights. The ports on the edge switches can be
pre-associated to storage ports, and control of the fan-in ratio (the ratio of the
input port to output port) can be established. With this pre-assigning technique,
the administrative team cannot overload any one storage port by associating too
many servers with it.
No fabric zoning Using no fabric zoning is the least desirable zoning option because it allows
devices to have unrestricted access on the fabric. Additionally, any device
attached to the fabric, intentionally or maliciously, likewise has unrestricted
access to the fabric. This form of zoning should only be utilized in a small and
tightly controlled environment, such as when host-based zoning or LUN masking
is deployed.