Software Manual

NAS Data Path Manager
13hp StorageWorks NAS Data Path Manager Installation and User’s Guide
As a general rule, when a non-optimal path is used, the request is going from one
HBA through a controller that communicates with the controller of the target
LUN. Therefore, the data request goes through two controllers to get to the
specific LUN, instead of just one. (This is true for all except XP controllers, for
which all paths are considered optimal.) See paths to Controller 2 in Figure 1 on
page 12. The overhead is the slight increase in time required for the first controller
to talk to the second controller. In reality the time that it takes is quite small for
each data transfer; however, it becomes significant when you multiply the extra
time by the number of transfers requested.
By default, NAS Data Path Manager assigns optimal paths as preferred paths, and
non-optimal paths as alternate paths. For XP arrays and some VA arrays, there are
no non-optimal paths. In such a case, no paths are assigned as failover paths by
default, so you must set them manually.
Note: A path can be either preferred or alternate, but not both.
Data Path Load Balancing
Load balancing allows the greatest throughput of data from an HBA to a LUN by
using all available preferred paths, instead of just a single path. This allows users
to have faster access to the data on the specified LUN. Without load balancing,
you are restricted to sending data over a single path, even when it is more efficient
to send data over multiple data paths. Lengthy wait times can occur as all data
requests queue up on the same path.
If one preferred data path fails, NAS Data Path Manager drops the failed path out
of the preferred path rotation, and continues to transfer data through other
preferred data paths without any loss of data. The switch-over is completely
transparent to applications, so normal operation continues without downtime.
All path events are logged in an Event Summary, allowing the administrator to
investigate and repair failed paths. A path monitoring daemon checks the health
and status of all paths every 10 minutes. If a failed path has been repaired, it will
be used again, according to the current load balancing policy.
Note: HA Package failover does not occur unless all preferred and alternate paths on a
single head fail.