HPP Enterprises P4000 SAN - Manual (Page 8)

8
Clustering and Network RAID
Since an individual storage node would represent a single point of failure (SPOF), the HP
StorageWorks P4000 SAN supports a cluster of storage nodes working together and managed as a
single unit. Just as conventional RAID can protect against a SPOF within a disk, Network RAID can
be used to spread a volume’s data blocks across the cluster to protect against single or multiple
storage node failures.
HP StorageWorks SAN/iQ, the storage software logic, performs the storage virtualization and
distributes data across the cluster.
Network RAID helps prevent storage downtime for XenServer hosts accessing that volume, which is
critical for ensuring that these hosts can always access VM data. An additional benefit of virtualizing
a volume across the cluster is that the resources of all the nodes can be combined to increase read
and write throughput as well as capacity
2 .
It is a best practice to configure a minimum of two nodes in a cluster and use Network RAID at
Replication Level 2.
Networking bonding
Each storage node supports multiple network interfaces to help eliminate SPOFs from the
communication pathway. Configuration of the network interfaces is best implemented when attaching
both network interfaces to an Ethernet switching infrastructure.
Network bonding
3 provides a mechanism for aggregating multiple network interfaces into a single,
logical interface. Bonding supports path failover in the event of a failure; in addition, depending on
the particular options configured, bonding can also enhance throughput.
In its basic form, a network bond forms an active/passive failover configuration; that is, if one path in
this configuration were to fail, the other would assume responsibility for communicating data.
Note
Each network interface should be connected to a different switch.
With Adaptive Load Balancing (ALB) enabled on the network bond, both network interfaces can
transmit data from the storage node; however, only one interface can receive data. This configuration
requires no additional switch configuration support and may also span each connection across
multiple switches ensuring there is no single point of failure to multiple switches.
Enabling IEEE 802.3ad Link Aggregation Control Protocol (LACP) Dynamic Mode on the network
bond allows both network ports to send and receive data in addition to providing fault tolerance.
However, the associated switch must support this feature; pre-configuration may be required for the
attached ports.
Note
LACP requires both network interfaces’ ports to be connected to a single
switch, thus creating a potential SPOF.
Best practices for network configuration depend on your particular environment; however, at a
minimum, you should configure an ALB bond between network interfaces.
2 The total space available for data storage is the sum of storage node capacities.
3 Also known as NIC teaming, where “NIC” refers to a network interface card