【请教】磁带库中装载单磁带机和多磁带机在应用上的区别

sun_up

IBM calls the features as DPF (Data Path Failover) and CPF (Control Path Failover) and supports it since 2002-3.

Automatic Failover
The automatic failover support provides error recovery on an alternate path when
a permanent error occurs on the primary path. This is transparent to the running
application. There are two types of Path Failover, Data Path Failover (DPF) and
Control Path Failover (CPF). They are closely related however the difference is that
DPF is an automatic failover support for the transfer of data, which provides error
recovery for systems connected to tape drives while CPF is an automatic failover
support for the transfer of commands to move tape cartridges. Below are examples
of different configurations that can be constructed.

Multiple HBA and Multi-Port Devices
Consider a multipath connection that consists of two Host Bus Adapters (HBA)
connected through a fabric to a multi-port device. By having two connections from
each device to the fabric, a configuration of four different physical paths can be set
up. e.g. HBA A to port 1, HBA A to port 2, HBA B to port 1, and HBA B to port 2.
One path is the primary path and the other three would be alternate paths. If the
primary path should have a failure, the Automatic Failover would try switching to
another configured path. Without path failover support, if a permanent path error
occurs (because of an HBA or cable failure, for example), the application fails.
However should that occur with this configuration there would still be two valid
physical paths for the data to be transmitted on and the running application
wouldn’t be affected.

Multiple HBA and Single Ports
For another example, consider the same multipath connection that consists of two
Host Adapters connected through a fabric to a single port device. This
configuration would supply two physical paths to the same device, however, if the
port or cable from the device fails the automatic failover would not work as that
connection is severed and a permanent path error occurs. If however the failure
was with one of the HBAs or their cables the automatic data path failover would
select the other HBA and the information would continue through the alternate
path.

Single HBA and Multi-Port Devices
Yet for another example, consider a single path from the HBA through the fabric to
a multi-port device. This also provides a failover path unless the failure is with the
HBA or the HBA’s cable, at which point the connection is severed and a permanent
path error occurs. Whereas, if the failure should occur on the device side there is
still an alternative path available for the information to go through.
At startup or configuration, the system detects multiple logical devices of the tape
drive. Each logical device is a physical path to the same tape drive. A backup and
restore application can open and use only one logical device at a time because they
represent the same physical device.

Without path failover support, if a permanent path error occurs (because of an
HBA or cable failure, for example), the application fails. It is possible to initiate
manual failover by restarting the application on the alternate logical device, but the
application has to be restarted from the beginning. A long back up or restore
operation may have been in progress when the path error occurred. Sometimes
manual failover may require operator intervention to reset the drive because a
SCSI Reservation could still exist on the failing HBA path.

When the path failover support is enabled on all logical devices, the device driver
configures them internally as a single device with multiple paths. The application
can still open and use only one logical device at a time. If an application opens the
primary device and a permanent path error occurs, the device driver initiates
failover error recovery automatically on an alternate path. If successful, the current
operation continues on an alternate path without interrupting the application. The
data path failover error recovery first restores the previous device state, SCSI
Reservation, and tape position, and then retries the failing operation.

For more details, refer to the document of "IBM Tape Device Drivers Installation and User's Guide"
at Page 12 https://www-304.ibm.com/support/ ... mp;uid=ssg1S7002972