The exceptional growth in the amount of data that companies must
store makes SANs (storage-area networks) one of the more important
information technologies today.
For companies where demands for storage are doubling every year,
the traditional approach of discretely allocating specific storage
devices to each application server is impractical.
Not only is scalability limited, but reliability and performance
considerations suggest that storage should not be confined to one
or two servers but instead become a network resource that can be
dynamically shared by multiple servers and applications. And
networked storage allows the distance between servers and media to
be extended, improving the physical security of data centres.
However compelling, these benefits of networked storage are only a
starting point. When storage becomes a network resource, it can
assume an unprecedented range of capabilities, including server
independence from media format, dynamic re-allocation of storage
volumes to server hosts (without service interruptions), and volume
mirroring both locally and remotely.
Thanks to these new capabilities, enterprise storage is freed of
some of its physical restrictions and becomes a plastic container
that can be quickly and easily shaped to satisfy changing business
requirements. Welcome to storage virtualisation, and to its promise
of more flexible and less expensive management of networked
storage.
Transcending physical storage
The goal behind storage virtualisation is to isolate the server OS
and its applications from the technicalities of storage devices,
thereby simplifying media management and the allocation of storage
space to each server - or host - computer. Normally a host computer
must manage specific storage devices with certain capacities and
other technical characteristics. With storage virtualisation
software in place, servers and their applications use logical
volumes that shield the host from the complexities of physical
devices.
A good analogy to storage virtualisation is RAID (Redundant Array
of Independent Disks). RAID solutions hide the technicalities of
the media, such as cylinder, track, and block geometry of a disk,
and present the OS with logical volumes. Storage virtualiSation
delivers similar functionality, but on a broad scale. Whereas RAID
is limited to a narrow range of storage devices attached to one
host, storage virtualisation extends the benefits of logical
volumes to the entire storage network.
The benefits of logical volumes are many. Without storage
virtualisation the host OS becomes strictly dependent on the
technical characteristics of the storage devices, so that
apparently simple activities - such as replacing or adding media -
require stopping the host and its applications while setting up new
storage devices and copying data. By contrast, storage
virtualisation removes most media management problems from host
computers and gives administrators the tools to create and connect
logical volumes to application hosts, regardless of the technical
specifications or network locations of the physical storage devices
involved.
A storage virtualisation system can create a logical volume using a
segment from a physical volume and can improve performance or data
reliability by aggregating multiple disks or segments. To further
ensure data reliability, logical volumes can be mirrored either
locally or remotely. Regardless of the logical volume's physical
composition, the host computer sees only an available data
container with a certain capacity; all other technicalities are
handled elsewhere using the virtualisation software as the
interface between host and media.
The obvious advantage of storage virtualisation is simplified and
less expensive storage administration. It is not surprising that
most storage vendors offer virtualisation capabilities, although
significant differences exist among solutions. A very apparent
distinction is the location of the storage virtualisation logic,
which can reside on host computers, on storage devices or on
network devices such as a switch, a router, or a dedicated
computer.
Moving beyond abstraction
These apparently contradictory approaches derive from an
architectural choice - often driven by the nature of the vendor's
networked storage products - that dictates where in the storage
network is more convenient to locate storage management software
and hence virtualisation logic. Currently there are no standards in
place to guide the implementation of storage virtualisation. This,
unfortunately, creates the battleground for animated and often
confusing debates, and makes purchasing decisions extremely
difficult.
No company makes technological choices in a vacuum; all must find
solutions that integrate with the existing infrastructure at
affordable cost. The variety of virtualisation offerings from
vendors such as Datacore Software, Softek, StorageTek, Veritas,
Vicom and Xiotech provides companies with a wider range of options
to meet their increasing storage requirements without disrupting
their existing platforms.
From a broader perspective, storage virtualisation could be the
"killer application" of networked storage and become one of the
crucial factors driving a company's choice of storage vendor.
Recent developments in this already active market suggest that
storage vendors are moving the focus of competition to this
area.
For example, Hewlett-Packard's recent acquisition of StorageApps
suggests that the company will reinforce its Federated Storage
Management Architecture with the storage virtualiszation technology
of StorageApps.
EMC, considered by many to be the market leader in networked
storage, has announced its intention to extend the storage
virtualisation capabilities of its Clariion and Symmetrix lines
with networkcentric, device-independent functionality that will
include simplified administration via companywide policies and
automated storage reallocation to meet performance
objectives.
Similar indications come from Compaq's revised ENSA-2 (Enterprise
Network Storage Architecture), which projects storage management
and virtualisation software such as SanWorks into a vision of
global storage networks. By contrast, Hitachi Data Systems is
pursuing a completely different approach, promising
interoperability between its hardware offerings and management
solutions from vendors such as Veritas and McData, and offering its
customers remote management of their networked storage.
It is difficult to resist making the analogy between networked
storage and networked computers, as they both serve the similar
purpose of connecting an application to its data repository. At the
dawn of the information era, network operating systems such as
Banyan Vines and Novell NetWare were instrumental to the popularity
of network computing because they offered companies undeniable
financial and practical advantages over the discrete administration
of data residing on PCs, despite the substantial cost of networking
equipment.
Storage virtualisation is likely to play a similar role in the
adoption of networked storage, by offsetting high hardware
acquisition costs with otherwise unattainable functionality and
administrative efficiency.
The bottom line
Storage virtualisation is the logical complement to networked
storage. By transforming a network of storage devices into a
virtual pool that can be shared among applications, storage
virtualisation solutions can deliver substantial financial and
practical benefits to companies facing a steep increase in storage
requirements.
Storage virtualisation technology is still evolving. Companies can
achieve immediate benefits, but must be careful when selecting from
the vast range of available solutions to choose those that are
compatible with existing investments and free of proprietary
restrictions.