Enterprise storage needs cannot be met with a single
hard disk or
server. With storage demands well into the
terabyte (TB) range, storage professionals rely on large groups
of disks to provide applications and data to enterprise users
through the corporate
SAN. A storage array provides an independent, standalone
platform where disks can be gathered into extremely large groups --
sometimes hundreds of disks. Once implemented, a storage array can
protect valuable corporate data by implementing
RAID features. An array can also contain a variety of drive
types, allowing businesses to organize their data into tiers.
Tiered storage matches drive cost, reliability and performance
against the value of corporate data. Storage array systems are
available across a wide range of capacities and features that must
be selected to meet the unique storage needs of an organization.
For example, array manufacturers typically offer systems for low,
medium and high-end deployments (e.g., their Celerra, Clariion and
Symmetrix DMX products respectively). In spite of their potential
complexity, however, there are some key ideas that can help bring
them into focus.
Understanding essential attributes
Simply stated, storage arrays are large boxes that hold lots of
hard disks, so it's important to select disks that meet your
storage needs.
Fibre Channel (FC) disks are common for mission-critical
installations where I/O performance and reliability are vital. Less
critical installations can often save substantial money by opting
for
SAS or
SATA disks instead. However, many storage arrays can
accommodate several different disk types within the same chassis
(e.g., FC and SATA), allowing for tiered storage. Regardless of the
disk types selected, the disks themselves are typically grouped
into RAID sets where
mirroring and
disk striping techniques can add redundancy and improve storage
performance.
Storage arrays must typically be connected to the corporate
storage infrastructure -- almost always a SAN -- including suitable
directors and switches. FC connections are most common, though
iSCSI installations are quickly gaining popularity in smaller
organizations. Array products must also support the protocols used
in the current storage infrastructure, such as
NFS,
CIFS or
FTP.
Storage arrays are also evaluated based on their capacity,
scalability and
cache size. Most arrays are purchased with an initial starting
capacity, but can typically be scaled up over time as storage needs
grow. For example, Sun's StorEdge 6920 base cabinet can grow to
16.4 TB (using 146 GB hard disks). Cache is little more than memory
used to hold data passed from the network until it can be written
to the array disks. Busy storage arrays should have more cache, and
systems like the StorEdge 6920 can scale cache from 2 GB to 28
GB.
Array management is critical
Huge storage volumes are certainly not available to every
network user or application. Storage administrators must
provision the capacity of each storage array, allotting storage
to applications, backups and other tasks. Consequently, storage
arrays must be "managed," and such management tasks often comprise
a majority of the total cost of ownership over time. Virtually all
management is handled through software, so administrators must pay
particular attention to any management software features. For
example, Sun's StorEdge 6920 includes data snapshot software, data
mirror software and data replicator software. In some cases,
provisioning and other management tools may need to be purchased
separately from the storage array, impacting the acquisition and
training costs for a new storage array implementation.