HP StorageWorks XPath OS 7.4.X Administrator Guide (AA-RVHDD-TE, February 2006)

XPath OS 7.4.x administrator guide 51
Proxy devices are presented to the fabric as being topologically attached to phantom domains created by
the FC-FC Routing Service. The MP Router creates two types of phantom domains for each edge fabric
accessed:
• Each EX_Port projects a unique front phantom domain (front domain).
• Each EX_Port also projects one translate phantom domain (xlate domain) for each edge fabric
accessed through it. All EX_Ports connected to an edge fabric use the same xlate domain number for a
remote edge fabric; this value persists across switch reboots and fabric reconfigurations. Xlate domains
are presented as being connected topologically behind one or more front domains. This allows
redundant EX_Ports or MP Routers with redundant paths to remote fabrics to present redundant paths to
proxy devices to an edge fabric.
Phantom domains are like logical switches that appear to be connected to an edge fabric through the
EX_Ports. The combination of front domains and xlate domains allows routing around path failures,
including path failures through the routers. The multiple paths to an xlate domain provide additional
bandwidth.
Connecting multiple EX_Ports to an edge fabric
You can connect multiple EX_Ports to the same edge fabric. The EX_Ports can all be on the same MP
Router, or they can be on multiple routers. Multiple EX_Ports create multiple paths for frame routing.
Multiple paths can be used in two different, but compatible, ways:
• Failing over from one path to another
• Using multiple paths in parallel to increase effective data transmission rates
Routing failover is automatic, but it can result in frames arriving out of order when frames take different
routes. The MP Router can force in-order delivery, although frame delivery is delayed immediately after the
path failover.
You can control whether in-order delivery is required; see ”Specifying frame delivery order” on page 28.
Source EX_Ports can balance loads across multiple destination EX_Ports attached to the same edge fabric
using exchange IDs from the routed frames as keys to distribute the traffic. This feature is enabled
automatically when the exchange-based trunking feature is enabled. See Chapter 8, “Using ISL trunking,”
for details on enabling this feature.
Matching fabric parameters
By default, the MP Router detects, autonegotiates, and configures the fabric parameters without user
intervention.
As an option, you can configure these parameters manually. Use the configure command on a switch in
the edge fabric to change the fabric parameters of a switch in the edge fabric. Then use the
portCfgExPort command to change the fabric parameters of an EX_Port on the MP Router.
For details see ”Supported modes” on page 59.
Fabric parameter settings must be the same on EX_Ports and on the fabrics to which they are connected:
E_D_TOV (error-detect timeout value), R_A_TOV (resource-allocation timeout value), and PID format. You
can set the PID format on an EX_Port when you configure an interfabric link.
The default values for E_D_TOV and R_A_TOV for an EX_Port match those values on HP StorageWorks
switches. You need to adjust these parameters for an EX_Port only if you have adjusted them for the fabric.
The default values for R_A_TOV and E_D_TOV are the recommended values for all but very large fabrics
(ones requiring four or more hops) or high-latency fabrics (such as ones using long-distance FCIP links).
SAN scalability
Fabrics have scalability limits; for example, the maximum number of Name Server entries allowed limits the
number of devices that can join a fabric. When you try to merge fabrics, this limit might be exceeded. But
using Fibre Channel routing, you need not merge the fabrics. Instead, you can selectively import only those
specific hosts or targets to be shared between the fabrics, and thus use Name Server entries more
efficiently.