No matter how much storage capacity you squeeze into a disk array,
it's just a matter of time until that space is completely filled.
Users everywhere are challenging existing storage resources with
applications that proliferate media-hungry data files. It's enough
to exhaust even the most generously sized
LUNs.
At the same time, growing concerns over regulatory compliance,
e-discovery and disaster recovery/business continuance are making
storage administrators reluctant to discard all but the oldest
content. Resulting storage shortages can impair application
performance and cripple mission-critical storage-intensive tasks,
such as database operations or transaction processing.
Careful planning can help administrators stay a step ahead in
the storage game. By using tiered storage tactics, they can move
noncritical data off expensive Tier 1 storage onto lower tier
disks, which are usually lower cost and higher capacity.
Comprehensive data deletion policies can ensure that data is kept
in the storage enterprise for the appropriate amount of time, then
safely eradicate the files that have met retention goals.
But no matter how well you manage your storage, you'll
periodically need additional storage. Here are eight best practices
for
disk array expansion or replacement.
Best Practice No. 1: Know the limits of your disk
arrays
You should know how many disks an array can hold, and how many
disks are actually installed at the moment. This is your "headroom"
for disk expansion. Small arrays are often filled to or near disk
capacity during the initial purchase. Because there's little
expansion capacity to work with, purchasing a new array is almost
unavoidable when you need more storage space. Midsized and
enterprise-class disk arrays usually ship with only a limited
complement of disks, allowing much greater headroom for array
expansion. Once you know the available headroom in each array, it's
easier to budget for additional storage chassis or make the
argument for forklift upgrades.
Connectivity is another limitation. Bottlenecks may develop at
the disk array as more users and applications demand storage
access. Again, small disk arrays offer limited connectivity, and an
infrastructure using them will require multiple storage systems in
order to distribute the storage traffic. Larger arrays offer
multiple Fibre Channel or Ethernet
(iSCSI) ports and can support far more
storage traffic. Larger arrays may also accommodate improved
connectivity through additional or faster port upgrades. Regular
performance monitoring will help track traffic bottlenecks in
and around the storage fabric.
Data migration is another critical aspect of tiered storage and
storage upgrades. Be sure to understand all the variables related
to data migration between your different storage systems, or
between disk types within the same storage system, and what the
impact will be on system performance. Administrators need to know
how quickly and efficiently their systems will conduct data
transfer.
Best Practice No. 2: Understand storage content and expand
storage accordingly
Disk performance and capacity can vary dramatically, depending
on the type of storage you need. You wouldn't want to buy
high-performance Fibre Channel disks to hold downloaded MP3 or .jpg
files. Nor would you trust critical database files to low-end SATA
disks. Use sound data classification techniques and tools to
identify data, estimate the importance of each data type to the
enterprise and move data to its corresponding class of storage or
storage tier. Once you implement data
classification, it's easier to forecast storage growth and
expand each corresponding storage tier accordingly.
Best Practice No. 3: Let disk prices fall
Volume discounts can be quite seductive. It's easy to see why
storage managers are tempted to purchase new disks in bulk, often
adding large swaths of storage to an existing array. However, disk
prices are falling so quickly that it's often cheaper to buy only
the storage needed in the immediatefuture. To makethemost of
plummeting disk costs, make smaller and more frequent purchases.
"Consider what you really need and buy that much, and maybe 10% to
20% more but not 50% more or 100% more," says Stephen Foskett,
director of data practice at Contoural Inc. By the time that
additional disk space is actually utilized, the cost of the disk at
that point will be lower than what you bought it for - - a waste of
storage capital.
Another issue with large disk space expansions is that the huge
block of additional capacity tends to be wasted storing data that's
not really needed. Storage managers get lazy and relax the
management criteria, and the added space fills much faster than
expected, leading to an upward spiral of larger and more costly
expansions. The added space also needs to be replicated and backed
up. Limiting disk space expansions can help ensure that management
processes are enforced.
There are three circumstances that justify overbuying storage.
1) If the total storage expansion is relatively small, the
incremental cost for more may be insignificant. For example, if you
only need another 1 terabyte (TB) of disk space, purchasing 2 TB is
relatively cheap. 2) When there is little, if any, regular budget
and storage acquisitions are part of related projects. Suppose you
need to roll out an ERP system that only needs 5 TB. If the project
budget allows for 15 TB to 20 TB and there is no other budget for
storage expansion, then the bulk purchase may be the only practical
means of adding significant storage in your organization. 3) When
the use or growth of storage is unpredictable, such as corporate
startups or multiple mergers.
If you decide not to overbuy storage, be sure to secure pricing
with your vendor for future purchases, not just for disks, but for
heads, controllers, cache or even professional services. "You want
to make sure that all that is locked in and predictable when you
take delivery of your new storage array," Foskett says. "Make sure
that cap is locked in for three years, four years or even the life
of the warranty."
Best Practice No. 4: Limit the number of vendors in your
storage environment
Unless you've inherited a lot of diverse equipment through
corporate mergers and acquisitions, or unless you need a feature or
function that's not available from your current vendors, it's a
good idea to limit the number of storage vendors in your
environment. Selecting similar equipment will ease interoperability
issues, reduce the proliferation of management tools and erase the
learning curve needed for configuration and maintenance. "One easy
way to do that is to select one vendor for your Tier 1 storage and
a second vendor for your Tier 2 storage," says Phil Goodwin,
president of Diogenes Analytical Laboratories Inc.
Best Practice No.5: Limit the number of tools in the storage
environment
Having fewer tools makes learning easier and allows more
cross-training and mutual support between storage professionals,
since specialization in particular tools is eliminated. Licensing
and update hassles are also simplified.
However, reducing the proliferation of tools is not always easy
(or even possible). This kind of simplification will depend on
heterogeneity between storage platforms. If you're stuck with
diverse storage systems, it may be impossible to avoid multiple
management tools until you have the opportunity to replace key
storage systems over time.
Best Practice No. 6: Move to larger disks
One way of expanding storage capacity while limiting the
proliferation of new disk arrays is to replace existing disks with
larger ones. For example, replacing 500 GB disks with 1 TB disks
will effectively double storage capacity within the same disk array
cabinet(s) without adding more cabinets to the data center.
However, the process of upsizing disks can be disruptive. Ensure
that the array can actually support the larger disks. In some
cases, the array may need a controller upgrade or firmware update
before larger disks can be installed. The actual replacement
process must also be choreographed so that data is properly backed
up and then restored to the new disks once they're installed and
configured. Many storage professionals opt to implement this tactic
in phases (e.g., one LUN at a time) over the span of weeks to
minimize the impact of downtime on the array.
Best Practice No. 7: Replace multiple arrays
simultaneously
Weigh the economics of replacing multiple arrays as opposed to
upgrading them. It's not just a matter of upfront cost. A new array
can offer benefits, such as added performance, data migration
capabilities, improved interoperability, better scalability and
reduced outages.
"In fact, many people will find that the cost of maintenance on
a four-year-old storage array is more than the cost of replacing
that storage array," Foskett says. You may find that replacing an
existing array outright is less expensive over time than upgrading
it, especially if you're already buying a new storage system from
the vendor.
Best Practice No. 8: Look for redeployment
opportunities
Decommissioning a disk array from one role does not prevent it
from assuming other storage roles in the enterprise. Yesterday's
Tier 1 storage system may make an ideal Tier 2 or Tier 3 storage
platform today, and many organizations mitigate their storage
investments by redeploying older storage arrays in less
performance-intensive tasks, i.e, backups and archives. Whenever a
disk array must be replaced, examine lower-tier storage needs and
see where the older array can be reused.
However, remember that older disk arrays may be out of warranty,
so the costs of maintenance and repair may make redeployment
impractical. In those cases, it may make more sense to scrap the
older array or leverage it in trade.