Storing Windows-based Oracle Databases on the HP NAS 8000

hp solution overview
Data integrity, reliability, scalability and high performance are top priorities in any database
implementation. The use of Network Attached Storage (NAS) devices to store database data files has
traditionally been discouraged because it was thought that network instability and perceived vulnerabilities
of non-local storage posed too great of a potential risk to data integrity and would greatly reduce system
performance. However, networking and storage technologies have improved and matured to a point
where what was once looked upon with suspicion is now taken for granted. While there are exceptions, it
is certainly no longer the rule that networks are too unstable for mission critical use. In fact, it is now almost
a given that in an office of any size, there will be some form of a computer network and probably some
form of non-local storage such as a file server. Whether this network consists of a couple of desktop
computers, a file server and a shared printer, or whether this network consists of thousands of systems
spanning cities/countries/continents, networks and some form of non-local storage are in widespread use
today. As network technologies have matured and improved, so have the protocols that utilize the network
topologies. At all levels, network protocols have become more robust, more efficient, and have helped to
create a more stable network infrastructure. Advances in storage technology, with high capacity hard
drives, fail over and high availability RAID Arrays, improved management and access techniques (to
mention a few) have also helped to reduce concerns about data integrity and storage performance. These
advances have opened the door for the use of non-local storage, specifically NAS, for many applications,
one of which is to store database data files. Additionally, and of growing importance, is the need for a
scalable solution that does not require system downtime in order to increase available storage. The HP
NAS 8000 has been engineered to provide a complete NAS solution providing data integrity, reliability,
high performance, scalability and ease of management, as well as heterogeneous file access.
The NAS 8000 makes extensive use of the concept of ‘virtualizing’ storage. This means the storage
subsystem is able to take all of its physical disk storage and present it as a pool of storage that can be
partitioned/divided up by the NAS 8000 administrator as each environment requires. The NAS 8000 uses
state of the art, HP technologies such as a special optimized operating system for fast and efficient data
access and an easy to use management GUI. The storage subsystem of the NAS 8000 is either the HP
Virtual Array 7100 (VA 7100), or the HP Virtual Array 7400 (VA 7400). The NAS 8000 can be
connected to the VA 7xxx Array either through direct connections, or through a SAN. The VA 7xxx storage
subsystem is designed to provide a very high level of data protection and reliability by utilizing dual array
controllers that provide redundant and concurrent access to array storage and implement HP AutoRAID
tm
technology. The dual controllers provide parallel data access pathways for all data located in the storage
subsystem; in the unlikely event that one controller should fail, the other controller will provide access to all
data in the storage subsystem. Each controller contains a battery to protect the contents of the non-volatile
ram (NVRAM) for at least three days in the event of a power failure. The dual controllers also provide
NVRAM mirroring through special hardware that guarantees that the mirrors are always in sync. The
mirrored NVRAM provides extremely safe and reliable storage for system configuration information and for
write cache. This massive redundancy within the VA 7xxx ensures that a single failure will not result in a
loss of data and usually will not cause a service interruption. Even if a catastrophic failure (such as a total
power failure for more than 3 days) were to cause a loss of the NVRAM contents, the system storage
configuration and storage maps can be recovered. At default/factory settings, the write cache is flushed to
the hard drives at a maximum of every four seconds. The actual time that data is staged in write cache
depends upon system configuration and utilization. The possibility of a single failure in the NVRAM is
extremely small and the VA 7xxx should not suffer any loss of data or interruption in service. The possibility
of a dual NVRAM failure resulting in data loss is significantly smaller. For performance, data protection
and data availability descriptions please refer to the HP Virtual Array 7100 and 7400 User and Service
Guide located off of the VA 7400 home page (see above for home page link).
One task that traditionally has been assigned to the database administrator is planning the storage layout
for a database. Until recently, conventional wisdom had the database administrator placing different
tablespaces/data files on different disk controllers/RAID arrays in an attempt to improve performance
through the concurrency allowed by multiple disk interfaces. Multiple hard drives/controllers allowed
multiple data pathways for ‘simultaneous’ access to data, as well as allowing multiple copies of critical files
to be stored on separate hard drives to protect against some media failures. Although this software
mirroring of critical data files offers some protection in the case of a media failure, it also uses precious
4