Quantum StorNext QD7000, Руководство по оборудованию

Страница: 22 / 141

point of failure. By creating an alternate path, you make sure that the drive shelves are accessible in
the event of a component failure.
For cabling examples, ranging from simple to complex, see
Drive shelf or drive tray cabling
on page
64.
Host cabling for redundancy
To ensure that the storage array remains accessible to the host in the event that a host channel fails,
establish two physical paths from each host to the controllers, and install alternate path software on
the host. This cabling topology, when used with alternate path software, makes sure that a redundant
path exists from the host to the controllers.
Attention: Possible loss of data access – You must install alternate path software or an alternate
path (failover) driver on the host to support failover in the event of a host channel failure.
For examples of redundant topologies, see Host cabling on page 50.
Cabling for performance
Generally speaking, performance is enhanced by maximizing bandwidth, which is the ability to
process more I/O across more channels. Therefore, a configuration that maximizes the number of
host channels and the number of drive channels available to process I/O maximizes performance. Of
course, faster processing speeds also maximize performance.
In addition to planning a topology that provides maximum performance, choose a RAID level that
suits the planned applications.
How best to approach zone configuration
Some of the host cabling topologies shown in this document require the use of a zoned switch. By
default, the switch uses no zoning, which is not sufficiently robust for most applications. You must
configure the switch before you use it.
When an initiator first accesses the fabric, it queries the World Wide Identifier (WWID) name server
for all of the attached controllers and their capabilities. Zoning is like a filter that the WWID name
server applies to the query from the initiator that limits the information returned by the WWID name
server to the initiator. A zone defines the WWID of the initiator and the WWID of the devices that a
particular zone is allowed to access. Devices that are not part of the zone are not returned as
accessible devices.
The fabric provides universal access for all initiators and targets. Any initiator can query (probe) the
fabric for all targets, which can affect performance when many targets are connected to the fabric.
The querying process also provides access to devices for which access is not needed. Use zoning to
limit the number of devices that an initiator can access. Within your storage area network, you should
zone the fabric switches so that the initiators do not “see” or communicate with each other.
Zone configuration is managed on a per-fabric basis. While you can administer zone configuration
from any switch, use the best practice of selecting one switch for all zone administration. Give
preference to the primary switches within the SAN, and choose only a switch that has the most up-to-
date storage management software and switch management software installed on it.
Host and storage array cabling for mirroring
The Synchronous Mirroring feature and the Asynchronous Mirroring feature provide online, real-
time replication of data between storage arrays in separate locations. In the event of a disaster or a
catastrophic failure at one storage array, you can promote a second storage array to take over
responsibility for computing services. See
Hardware installation for Synchronous Mirroring and
Asynchronous Mirroring on page 109 for detailed information.
Cabling concepts and best practices | 13