This is an excerpt from the bestselling book
Oracle Grid & Real Application Clusters. To get immediate
access to the code depot of working RAC scripts, buy it
directly from the publisher and save more than 30%.
Multipathing generally allows
two or more data paths to be used simultaneously for read/write
operations. This enhances performance by automatically and equally
dispersing data access across all the available paths. Many storage
and storage software vendors offer this kind of reliability.
Some of the widely used
solutions such as EMC?s Power Path, HP?s Secure Path, and Veritas?s
Dynamic Multipathing solution will be examined next.
Power Path (Dell / EMC)
EMC?s Power Path helps to
improve the high availability of the data or I/O path and manages
the heavy storage load. It automatically detects host failures and
recovers storage. Power Path is host-based software that runs on
UNIX, Windows, and Linux servers. It can manage the protocols in
both SCSI and fiber physical interconnects. Power Path creates a
virtual power device that provides failure resistance and a load
balanced path. Power devices are virtual objects that contain file
systems or raw partitions. They are managed by a volume manager or
host system. Each power device represents a LUN at the host level
but remains available through multiple I/O paths. The detailed path
is shown in Figure 5.11.
Figure 5.11: Power Path as
implemented in EMC Clariion storage array
HP's Secure Path is multipath,
high availability software that manages and maintains continuous
data access to HP storage systems. There is no single point of
failure from the server to the storage. Secure Path is host resident
software that monitors the data paths between the server and the
storage to increase the availability of the information. In the
event a path failure is detected, Secure Path fails over the data
path to an alternative path. When the original path becomes
available, Secure Path can automatically fall back to the original
path. Secure Path can also balance the workload among available
paths to optimize system performance.
Dynamic Multipathing (DMP) is a
feature of the Veritas Volume Manager. It offers greater reliability
by providing a path failover mechanism. In the event a connection to
a disk is lost, the system continues to access the critical data
over the remaining sound connections to the disk. DMP also provides
greater I/O throughput by balancing the I/O load uniformly across
multiple I/O paths to the disk device.
PolyServe Matrix Server supports
multipath I/O in its driver stack. MxMPIO offers solid redundancy in
the event of an HBA failure. Additionally, MxMPIO offers I/O load
balancing by multiplexing I/O for varying LUNs between the different
HBAs in the system. Bringing this feature to the Intel-based
server-clustering arena makes commodity-based datacenter solutions
one step closer to reality.
The database is constructed and
maintained on file system files or raw partitions, which are
entities at the operating system level. They are directly attached
to storage devices, NAS devices, or SAN volumes. There are several
different layers of abstraction between the physical disks and the
objects that the database uses. The physical drive is at the lowest
level, next comes the LUNS (logical unit numbers) of manageable
sizes employing an appropriate RAID level. LUNS are presented to the
host system as physical objects.
The Logical Volume Manager (LVM)
is a software module running at the host level and manages the
physical objects and presents them as logical units that the
application can use. The volume manager hides the physical
attributes of the disks or LUNs by introducing virtualization. The
volume manager can also execute the RAID operations if needed.
Within the volume manager, storage devices are grouped into disk
groups (e.g. Veritas disk group).
Within a disk group, a logical
volume can be created with an optional or desired RAID level.
Finally, the logical volume is either presented as a file system
after mounting, or presented as a raw partition for application use.
The VM tool provides a very flexible way of managing the volumes and
file systems. Without a volume manager, LUNS are physical disks with
Figure 5.12: Volume
Configuration using Veritas Volume Manager
Thus the volume manager hides
the details about where data is stored in the hardware from the
entire system. Volume management allows the editing of the storage
configuration, for example, setting up a software RAID or extending
volume size, etc., without actually changing anything on the
hardware side. Some volume manager tools allow snapshot copies to be
taken of volumes to move or backup. Thus, for either creating the
raw partitions (or volumes) or setting up of file systems, the
volume manager is a handy tool.
Figure 5.12 above shows how the
LVM provides volumes and file systems to a host. LVM hides all the
complexity in the backend of the disk drives or the storage units.
The database system will just interact with the volumes presented by
The Veritas volume manager has
been in use on many platforms like Solaris, HP, and AIX. Now,
Veritas even has the Volume Manager to support Windows systems. On
the Linux side, SuSE LVM has been in use for a long time.
However it is important to note
that cluster volume management is different from the normal volume
management. Cluster Volume management provides the volumes for all
the nodes in the cluster infrastructure.