|
So what is a Storage Area
Network?
A Storage Area Network, in its purest sense, is a separate
computer network, typically based on a 'fabric' of fibre
channel, switches and hubs that connects storage devices
to a heterogeneous set of servers on a many-to-many
basis. A SAN can also enable direct storage-to-storage
interconnectivity, and lends itself exploiting new breeds
of clustering technology and to get the best out of
Network Attached Storage devices that can intelligently
provide disk and tape capabilities to one or more servers.

The technologies that constitute a SAN are as follows:
- The critical enabler is fiber technology. This
is the ability to send data across fibre optic channels
at very high speeds and bandwidths to sites as far
away as 10 kilometers.
- Fiber enabled servers, disk arrays and other intelligent
storage and other devices are then connected to the
'fabric' by fiber through sophisticated switches and
hubs. These may be configured in highly scalable manner
to provide access from hundreds of servers in many
locations to hundreds of terabytes of shared storage
resources attached directly to the Storage Area Network.
- Zoning and other security facilities are emerging
to enable the same levels of security that we have
come to demand on the mainframe. What these technologies
do is enable the removal of storage devices from their
servers and make the storage generally available across
the network on a 'many to many' basis.
In its simplest form SAN simply
provides server to storage access across fiber - a 'SCSI
on steroids' capability. In a more sophisticated form
SAN enables a disk or tape array to be accessed by two
or more servers at high speed across the fibre channel.
The benefits of this include Improved
Performance, Greater Connectivity and the elimination
of redundant data. A Storage Area Network also expedites
the use of advanced clustering solutions thereby enabling
new levels of availability and business continuity.
A SAN also enables us to relocate backup, restore, file
migration and replication of data from the servers and
local/wide-area networks and have direct data movement
from disks/tapes to other disks/tapes across the SAN
fiber.
The benefits here include:
- backups done quietly behind the scenes server power
is freed up for the business applications network
capacity is released for the users
- remote disaster recovery sites up to date - to the
minute!
All this without having to change
existing applications, database management systems or
the way the user connects to their applications via
their current local area network - we are simply moving
the storage away from the servers onto its own network.
And the applications will be delivering 'local' speed
of access to data on the SAN where the data may in fact
be many kilometers away.In other words, a SAN represents
the removal of storage from their servers and makes
the storage generally available across the network.
An alternative and
interesting way to visualize a Storage Area Network
is shown below. The Storage Area Network
is shown within the center. This is where the
corporate data resides - a pool of shared on-line, off-line
and intelligent storage devices made available via the
fiber fabric.

Users connect via a Local Area
Network (LAN) and/or Wide Area Network (WAN) through
the application servers into the data. In this
way we can see that the application server is simply
acting as an access path into the data. This is
a new way of looking at our business-computing environment.
Rather than connecting to, for example, the email server
we are accessing our email data through the server.
Further, if we have a suitable architecture, then should
the application server fail, compatible spare server
capacity can take over the user processing needs easily.
This is a crucial point to note - this new storage topology
facilitates a new architecture for enhancing availability,
and leads us to the interesting notion that application
servers in a SAN-aware cluster can be considered as
'Application Server Peripherals' around the users' data.
It also shows that the typical 'failover' high availability
solution of today that requires 100% redundant hardware
can be replaced by one where, say, a 32-machine cluster
can enable each server to provide 'application takeover'
facilities for its peers. This provides equal or better
availability with far less hardware. This
new architecture provides a number of significant benefits:
Increased
Flexibility |
| Isolated
data sources can be interconnected and made generally
available to multiple servers. The limit for
how far apart these sources can be is defined by
the enabling infrastructure of the fibre channel
itself. This facilitates the re-use of storage
in a more efficient manner than ever before. Indeed,
through the SAN architecture, we can consider allocating
storage from one application to another - even though
the applications are running in different physical
locations. The result is a far better return
on investment from our existing infrastructure.
|
Improved
Scalability |
| Servers
and their associated storage are no longer tightly
linked together. New storage devices can be
dynamically added to the central pool of storage,
without having to add a new server. In a similar
way, additional servers can be dynamically added,
should we require additional processing power for
our applications. |
Improved
Performance |
| High
performance access to global data is achieved because
of the inherent performance characteristics of the
enabling fibre channel technology. In addition,
the SAN architecture makes it possible to relocate
various activities such as backup, restore, file
migration and replication of data. Instead
of passing data across the local/wide-area networks
via servers, we have the opportunity to move the
data directly from disk/tapes to other disks/tapes
across the SAN fiber. See (ii) in Figure 1.
In this way, server power is freed up for the business
applications and network capacity is released for
the users. |
Improved
Availability |
| The
Storage Area Network dramatically improves availability
of business processing. For example, any server
can take over from a failed applications server
- as they share access to the same storage and users.
We are no longer constrained by bottlenecks in terms
of how we get to our data - we can now reach our
corporate data through any application server because
of our improved clustering solutions. In addition,
data can be automatically replicated to where it
is needed in business terms. Technology is now becoming
available that will enable such replication to occur
at either the disk/volume level or at the database/file-system
level. For the really critical applications a change
of data value can be updated on all replicas simultaneously
- synchronous replication. For most data, however,
it is perfectly acceptable that replicas are a few
seconds or minutes out of date to their master copy
- asynchronous replication. A fundamental utility
within the Storage Area Network is a set of replication
facilities to meet these needs. These replication
facilities must be completely accurate, reliable
and exhibit very high performance and thereby give
the company confidence to remove all the 'private'
copies that probably exist to date.
Hence the Storage Area Network model provides inherent
availability of data by the use of automatic data
redundancy, automatic backups and the maintenance
of nearby disaster recovery copies. User-level replication,
as mentioned above, also adds further resilience
if provided. Clustered servers with shared access
to the data that can dynamically switch users (and
applications) between peers will dramatically improve
user and business application availability. |
Improved
Manageability |
| In
many senses Storage Area Networks lend themselves
to improved management by their inherently better
reliability and 'centralization'. Whereas
previously the storage was "hidden" behind specific
servers, it is now generally available. This
is inherently safer and simpler to manage - we can
now deal with the two separate concepts as such.
Automated monitoring and management tools can identify
faults, isolate the cause and automatically fix
them - without human intervention and with imperceptible
effect on the service. Other than data
redundancy for resilience purposes, the Storage
Area Network enables duplicate data to be removed,
saving on total disk capacity needed. Utilities
like backup, HSM and replication can then be carried
out within the storage pool and any optional Network
Attached Storage devices. From a storage perspective
things are now much simpler, assisting in ease of
use and centrally managed storage policies. |
Improved
Reliability |
| Inherent
within a SAN structure is RAID disk and tape technology
for both improved reliability and performance.
Fibre channels and sophisticated switched fabric
technology ensure no single point-of-failure on
the network component. Further, over time, SAN technology
will provide the ability to replace or repair any
component during normal operation, reconfigure the
system, add new components and otherwise enable
the vast majority of changes to be made on the fly.
|
All of these points provide quantifiable
business benefits relating to increased "Return on Investment"
and enable the IT function to deliver on, or indeed
to exceed, required Service Level Agreements.
We can now view a distributed set of storage devices
as though they were one single homogeneous mass.
Hence, if a given application requires additional disk
capacity, then we can re-assign disk capacity from any
other application on the SAN - even though these other
resources may be located many miles away from the primary
application. This gives us an enormous improvement
in "Return on Investment" because we can re-use storage
devices that previously were hidden behind a given server.
So what does this do for
my business?
Why should the president or managing director of a company
find storage area networks even vaguely interesting?
How will it improve a company's financial or other performance?
Storage Area Networks may enable some companies to operate
in a business sense in a radically new way, or in a
way they have aspired to in recent years but the technology
had really not been up to the challenge.Such companies
and government organizations are typically geographically
dispersed and depend upon timely and accurate information
as a primary corporate asset. In these organizations
data is the life-blood of the business. Today
the data required is stored in multiple, isolated locations.
- Unmanaged data duplication means inaccurate information
on which to base business decisions.
- Isolated data means that the 'total picture' cannot
easily be ascertained
- Timely information is often impossible to assemble
when required.
The information required by these
companies may be contained in huge data warehouses,
but is more likely to be in the form of a combination
of widely dispersed sources. These might include data
feeds from outside the company, Intranet web sites,
multiple business applications and other databases,
email servers; but the vast amount of information is
still held as normal files on the file systems. To enable
businesses to operate with a wide-open shared data concept
requires systems to have high performance access to
data for all authorized users across the business.A
Storage Area Network can enable isolated data sources
within say a 30 km area to be interconnected and made
generally available. Within a Storage Area Network
or across one or more interconnected SANs, a most useful
capability is automatic controlled replication of critical
corporate data to where it is needed in business terms.
This may be for a disaster recovery site or to provide
local copies for performance reasons. Such replication
needs to faithfully reflect the master copy, albeit
a few seconds or minutes later. An example might be
a corporate Intranet web site with an employer address
book that everyone can use in replacement of out-of-date
photocopies. And every time you connect to your network
your private copy is synchronized on to your computer.
Imagine that for corporate data across subsidiaries
within a global company. Another business requirement
is that the information is always available. 'Always'
does not mean 99.999% system availability - it means
100% availability by users and their applications to
the data. A Storage Area Network provides inherently
very high availability by the use of automatic data
redundancy, automatic backups and the maintenance of
nearby disaster recovery copies. Clustered servers with
shared access to the data that can dynamically switch
users and applications between peers will dramatically
improve user and business application availability.
On the fly change of all components is also a promise
of SAN - whether that be a replacement disk that is
about to fail, a server that needs maintenance, a software
upgrade or a part of the network that is to be enhanced.
The SAN architecture should also assist in providing
a smoother more consistent service by adding load balancing
to the high availability and performance already mentioned.
The inevitable conclusion
Many companies will evolve towards SAN technology without
really realizing it, as new disk arrays, offline storage
devices, servers and software are all made fiber and
SAN connectable. Others will adopt Storage Area Networks
aggressively to gain the potential benefits of far better
utilization of resources, new levels of availability,
performance and the ability to better share corporate
data across their company. Earlier adopters of Storage
Area Networks must select proven suppliers with a pedigree
for producing reliable hardware or software. They must
advance aggressively to gain benefits of a SAN, but
with due care and planning. Management tools and suitable
'SAN aware' utilities should be deployed early to help
control this new 'unknown'.
Storage Area Networks will take
several years to mature during which we will see these
early adopters work through an evolution from 'SCSI
on steroids' to an environment with outstanding availability
and performance for global applications and globally
shared data. The key to success is the establishment
of a 'storage centric' environment supported by a high
performance low latency fiber fabric that provides users
with highly available access to clusters of application
servers with many-to-many connectivity to shared online
and offline storage.
|