White Papers

4
high availability and performance load balancing across multiple network ports,
allowing the iSCSI Initiator software on the server to utilize multiple network interfaces
with a single iSCSI target. Refer to Figure 2 for a visual. The multiple paths can be
utilized in one of two ways, either to improve the throughput between the initiator and
the target or to implement redundancy for high availability. The multipath policy set
on the initiator (SLES server) determines how the multiple connection paths are
utilized.
Figure 2: Multipath I/O connection environment
SUSE Linux Enterprise Server supports ‘round-robin’ and ‘failover’. A multipath policy of
‘round-robin’ will utilize multiple active connections between the iSCSI initiator and a
target. A policy of ‘round-robin’ in the configuration shown in Figure 2 would double
the potential network throughput between the iSCSI initiator and target. When a
multipath policy of ‘fail-over’ is chosen, only a single connection will be active
between the iSCSI initiator and target. Therefore the network throughput between the
initiator and target is not increased. Instead, a redundant connection path is available
to be activated in the event of a connection failure. If the multipath daemon on the
SLES server detects a failure in the primary connection path, it will redirect traffic to the
alternate connection path.