Brocade Fabric OS Encryption Administrator's Guide v7.1.0 (53-1002721-01, March 2013)

298 Fabric OS Encryption Administrator’s Guide (SKM/ESKM)
53-1002721-01
Best practices for host clusters in an encryption environment
5
The fan-in ratio for a target can be higher depending on the maximum bandwidth accepted by the
target. If the I/O throughput across all initiator ports accessing the target port is well balanced, it is
recommended that the maximum fan-in ratio be kept to 8 Initiator ports to 1 target port for optimal
performance. Note that this recommendation holds for initiators running at 4 Gbps or less. If a mix
of 8 Gbps and other 4 Gbps or less initiator is used, then the maximum fan-in will depend on the
maximum sustained bandwidth these initiators would be pushing together over the link to the
same target port and across all the target ports hosted on a given encryption engine.
NOTE
If the performance license is not installed, 48 Gbps of full duplex encryption bandwidth is available
on the encryption engine, Each of the six encryption blocks will use two ports instead of four,
reducing the fan-in ratio by a factor of two.
Best practices for host clusters in an encryption environment
When host clusters are deployed in a encryption environment, please follow these
recommendations:
• If two encryption engines are part of an HA cluster, configure the host/target pair so they have
different paths from both encryption engines. Avoid connecting both the host/target pairs to
the same encryption engine. This connectivity does not give the full redundancy needed in
case of encryption engine failure and failover to another encryption engine in an HA cluster.
• For Windows-based host clusters, when a quorum disk is used, the quorum disk plays a vital
role in keeping the cluster synchronized. It is recommended that you configure the quorum
disk to be outside of the encryption environment.
• For AIX-based Power HA System Mirror host clusters, the cluster repository disk should be
defined outside of the encryption environment.
HA Cluster deployment considerations and best practices
It is mandatory that the two encryption engines in the HA cluster belong to two different nodes for
true redundancy. This is always the case for Brocade Encryption Switches, but is not true if two
FS8-18 blades in the same DCX Backbone chassis are configured in the same HA cluster. In Fabric
OS v6.3.0 and later releases, HA cluster creation is blocked when encryption engines belonging to
FS8-18 blades in the same DCX Backbone chassis are specified.
Key Vault Best Practices
• Make sure that the time difference on the Brocade Encryption Switch and the SKM/ESKM key
vault does not exceed one minute.
• When encrypted disk LUNs are to be configured or moved to an Encryption Group that uses a
different key vault, make sure to decommission the encrypted LUNs from the old Encryption
Group.