NAS server to another, the filesystems remain accessible without remounting and,
most importantly, processes with active file handles accessing files in the NFS
filesystems are not impacted during server transitions.
The VNFS technology is important for two reasons:
• Transition rather than failover. If a HP Scalable NAS server fails, the Virtual NFS
Services running on the failed node will be transparently transitioned to a backup
NAS server. The NFS clients and the processes with open file handles on the
filesystems involved will not be affected in any way.
• Rehosting. Using the HP Scalable NAS Management Console or CLI, the admin-
istrator can proactively move a mounted Virtual NFS Service from one NAS
server head to the other for such reasons as load balancing or maintenance. This
operation is also fully transparent at the NFS client level and there is no need to
stop applications. Just a simple GUI drag and drop.
The benefit of this technology in an Oracle deployment should be quite clear. With
the failover capability, a single point of failure due to a server failure is averted. And
in the case of a failover event, there is no need to remount a filesystem holding Oracle
files, so there is no need to stop and restart the database, which increases application
availability and avoids lost transactions.
In the event a NAS head fails, all filesystems actively presented by this server rapidly
and transparently failover to configured backup NAS heads. This means that, for
example, an Oracle query running at the time of the failure will continue as though
nothing happened. In a RAC environment, if a NAS head presenting datafiles should
fail, there is no global RAC meltdown. With RAC there is one copy of the database
(known to Oracle, at least), and suffering a “sever and reattach” sort of NAS head
failure will impact every instance of the RAC cluster. All of this is avoided. Service
interruption due to down NAS servers can happen only if every server configured in
the cluster to support a VNFS service is down.
This multi-headed NAS approach is called symmetrical because there is no one device
that “owns” a particular filesystem, and there are multiple NAS heads that can present
it simultaneously. Each filesystem can be presented by any of the active NAS heads
in the cluster at the same time—not just in a failover situation.
Since there are no ‘”owners” of filesystems, and since any server can present any
filesystem, as more and more servers are added, there is higher availability offered
and more resources to service requests. There is no need to physically copy hot data
to different filesystems in order to spread the workload across over-burdened NAS
heads. It goes beyond clustering paired single-headed filers in that many more than
two servers (in fact, up to 16) can be used to present any filesystem. Of course, HP
Scalable NAS does support NLM locking to prevent accidental sharing of single
instance datafiles by two Oracle servers simultaneously.
HP Scalable NAS File Serving Software provisioning guide for Oracle
11

---

...

Otros modelos de este manual:
datos de dispositivos de procesamiento - HP DL380 - WSS Clustered Gateway (1.56 mb)
datos de dispositivos de procesamiento - HP DL380-SL Clustered Gateway (1.56 mb)
datos de dispositivos de procesamiento - HP ProLiant DL380G5-WSS Storage Server Clustered Gateway (1.56 mb)
datos de dispositivos de procesamiento - HP ProLiant DL380G5-WSS Storage Server Initial Cluster (1.56 mb)