These days, the storage
demands of users and applications are spiraling out of control in
many enterprises. Additional
disks and storage subsystems are frequently
implemented to meet demands. But increased storage resources
carry a heavier management penalty -- storage administrators
must remember which applications are tied to what storage,
allocate sufficient storage to address future growth and
manually track performance. Over time, this inefficient manual
process leads to wasted storage space.
@28884 Storage
virtualization alleviates these traditional
problems by implementing a layer of abstraction between
applications and physical storage, allowing storage to be
combined and treated as a ubiquitous resource, regardless of
location. While
storage can be virtualized in a variety of
ways, the ultimate objective is to improve storage utilization
(reducing or forestalling capital expenditures) and enhance the
effectiveness of storage management. Aside from considering the
obvious issues of pricing and support, here are some other
important points.
Understand what storage virtualization must bring to the
enterprise. Storage virtualization offers numerous potential
benefits to the enterprise, but it's important to identify the
benefits that are needed by your particular organization. Improved
storage utilization, capital cost savings and ease of management
are the three most common benefits, but there are other advantages
that appeal to storage administrators. For example, virtualization
can ease disaster recovery (DR) or
business continuence planning by allowing
nonidentical hardware between sites. It can handle data
migration between storage platforms,
tiers and sites. Virtualization also eases
storage capacity expansion, often automating many of the manual
tasks needed to allocate storage to applications when
needed.
Weigh the added complexity. While storage virtualization
can ease some complexity problems, it can also add others. Many
virtualization products require additional hardware or software in
the infrastructure. Beyond new servers or
switches, determine what host device
drivers, path managers,
agents or shims are needed to support the
prospective virtualization product. Maintenance is usually first
to feel the pinch here -- it's easy for IT staff to become
bogged down with patching and updating a myriad of storage
virtualization servers when hardware is replaced or new versions
become available. Inadequate attention to maintenance can result
in version disparity leading to stability and performance
problems. Evaluate any storage virtualization product from a
management and maintenance perspective, and determine if the
problems that it solves are outweighed by the new issues that it
introduces.
Know where storage virtualization best fits in your
environment. There are typically three means of implementing
storage virtualization; host-based, array-based and fabric-based
virtualization. Host-based virtualization relies on software,
installed on host servers, which monitors data traffic and storage.
Veritas Storage Foundation from Symantec Corp. is an example of
this type of product. Dedicated appliances, such as the File
Director 7200 appliance from NeoPath Networks Inc., follow a
similar direction. Array-based virtualization integrates the
technology directly into the storage array itself, such as a
TagmaStore array from Hitachi Data Systems Inc. (HDS). More
recently, there is growing attention paid to fabric-based
virtualization that runs dedicated software on intelligent switch
devices. Each approach offers unique advantages and disadvantages
that can impact its performance, scalability, cost and
reliability.
Consider the scalability. There is a finite limitation to
the amount of storage that a virtualization product can support.
Understand the tradeoff between scale and performance. This is
particularly treacherous because many virtualization initiatives
begin as test or pilot deployments before broader deployment
through the enterprise. Consequently, scaling issues may not appear
until later in the deployment cycle. An up-front evaluation of
scaling can help to weed out unacceptable products and allow
network administrators to plan for future infrastructure
updates.
Consider the interoperability. The promise of
cross-vendor storage utilization has been a compelling feature of
this technology, but true
heterogeneous support is still lacking in
virtualization environments. For example, array-based
virtualization typically locks in the array vendor. Host-based
and fabric-based virtualization products also impose a certain
amount of vendor lock-in with the software or appliance that
embeds the software. Virtualization adopters should closely
investigate potential products to determine their compatibility
within the current environment. Also, consider compatibility
with likely upgrades or updates into the future.
Test and start small. Analysts typically recommend a
thorough lab evaluation of any storage virtualization product prior
to any purchase commitment -- including a comprehensive review of
decommissioning drills. Once a purchase decision is actually made,
the best advice is to start implementation on a small scale and
then build out the virtualization systematically. This conservative
approach allows ample time for administrators to become accustomed
to virtualization management and prevents unforeseen oversights or
problems from crippling an entire data center.
Understand how to undo the implementation. Storage
virtualization isn't perfect. Performance issues, scalability
limitations and interoperability problems are just a few reasons to
decommission a virtualization product. Similarly, an organization
may choose to discontinue one product in favor of a more
appropriate product. Unfortunately, backing out of virtualization
is extremely disruptive to applications -- and typically confusing
to administrators. Before committing to a virtualization product,
discuss any back out options with the vendor. ***
Return
to the beginning