| Risk: In a SAN environment, we potentially have data with differing levels or need-to-know stored on the same "system". A high level of assurance that a valid entity (user/system/process) of one set of data is not inadvertently given access to data that is unauthorized. Depending on the data and implementation, lack of hard zoning could provide access to classifed, administrative configuration, or other privileged information.
A zone is considered to be "hard" if it is hardware enforced. In other words, it is considered “hard” in that they are always enforced by the destination ASIC. "Soft" zoning is more flexible but is also more vulnerable.
In "soft" or WWN-enforced zoning, however, the HBA on the initiating devices store a copy of the name server entries, which were discovered in the last IO scan/discovery. It is possible for the HBA to include old addresses, which are no longer allowed in the newly established zoning rules. So your goal is to mitigate this risk in some way.
If hardware enforced zoning is used this is not an issue as the destination port will not allow any access regardless of what the OS/HBA “thinks” it has access to.
Supplementary Note: Registry State Change Notifications ( RSCN ) storms in large SAN deployments are another factor of which the system administrator must be aware. RSCNs are a broadcast function that allows notification to registered devices when a state change occurs within a SAN topology. These changes could be as simple as a cable being unplugged or a new HBA being connected. When such changes take place, all members would have to be notified of the change and conflicts would have to be resolved, before the name servers are updated. In large configurations it could take a long time for the entire system to stabilize, impairing performance. Effective zoning on the switch would help in minimizing RSCN storms, as only devices within a zone would get notified of state changes. It would also be ideal to make note of business critical servers and make changes to zones and fabrics that affect these servers at non business critical times. Tape fabrics could also be separated from disk fabric (although this comes at a cost). Statistics of RSCN's are available from a few switch vendors. Monitoring these consistently and considering these before expansion of SAN's would help you with effective storage deployments. |
