Specifications
© IBM Copyright, 2012 Version: January 26, 2012
www.ibm.com/support/techdocs 30
Summary of Best Practices for Storage Area Networks
For systems in high availability SAN environments, individual hosts will have multiple
paths to their respective storage volumes. Without multipathing utilities, many operating
systems would treat each path as a completely separate and unique resource even
though the end storage resource was the same in all cases. Multi-pathing
utilities/functionality are typically inserted into the IO path between the physical HBA’s
device driver and the operating system’s kernel, or logical volume manager, to “hide”
the details of multiple paths and only present one logical instance of each storage
volume. Another purpose of these utility applications is to silently handle a path failure
while normal IO operations continue.
However, multi-pathing can also experience the situation of “too much of a good thing”
by having too many paths between a host and a storage resource. As the number of
paths increases, the load on the CPU by the multi-pathing application can increase
significantly. One misunderstood fact is that doubling the number of paths between a
host and a given storage volume does not mean the total usable bandwidth has also
doubled. This observation is due to the computing needs of multipathing algorithm as it
makes use of the additional paths.
One condition which has been observed too many times within the SAN Central team is
a lack of coordination between server and SAN/storage administrators concerning HBA
maintenance. The historical experience has been a matter where the server
administrator considers HBA firmware updates as a function of the SAN/storage
administrator since the HBA is part of the SAN. Conversely, the SAN/storage
administrator considers HBA maintenance activities to be in the purview of the server
administrator since the HBA is located within the server. The key point and best
practice guideline is that the administrators need to arrive at a mutually agreeable
means where one, or both, are responsible for routine HBA maintenance actions.
8 Change Management
“Change Control” is one of the most discussed concepts in the management of
Information Technology. There are entire international standards dedicated to defining
it, and very expensive and complicated software created to implement it. However, in
practice, change control often ends up being implemented in ways that do little to
actually make the environment operate better.