Simplifying the Tiers

Optimizing on storage environments is what every CIO wants to achieve. Many
enterprises are looking at ways and means of optimizing their storage assets;
while many have taken the virtualization route that is consolidation through
virtualization is an important first step towards realizing many of these
benefits. So, it becomes paramount in effectively configuring and managing the
virtualized tiered storage environments.

One of the most pressing issues facing organizations today is the need to
reduce storage costs. Simultaneously, single vendor management tools and
isolated pools of storage fragment the infrastructure, making it difficult to
share resources, improve operational efficiency. By implementing virtualized
tiered storage architecture, an IT organization can dramatically improve storage
capacities and lower capital and operational expenses. Virtualized tiered
storage better aligns data on storage systems by allowing users to efficiently
match storage attributes with the service level needs of individual business
applications. To achieve a seamless virtualized tiered storage environment, it
has to start from the design phase itself.

Configuring Tiered Storage

There are some key factors enterprises should consider while configuring,
managing and designing the tiered storage. When we look at design elements
oriented towards higher performance, the key things one should look at are
response time and throughput. The finer elements that need to be studied here
are aspects like I/O operations per second per gigabyte and megabyte capacity.
The second design element should factor in the critical two
parametersreliability and availability. So, the things to be studied here are
aspects like the extent of modular storage and whether its sufficient, and the
kind of RAID required. The major thing that needs to be considered is the
Logical Unit Number (LUN) and the kind of sizes required. The final design
element is the capacity management and the ability to scale the tier in the most
efficient manner.

The underlying design must facilitate certain key and critical storage
deliverables. First, such deliverable is the storage performance. As we look at
the individual parameters that impact the storage performance, a major downer
that ushers in inefficiencies comes from the storage being not tuned to the
demands. If the storage is not tuned to the demands it can dramatically erode
any benefits derived from virtualization. Storage needs to respond quickly to
increased workloads and overcome I/O bottlenecks automatically. But,
administrators do not have the time or resources to constantly tweak the
performance characteristics of their storage systems. As overall workloads
increase, storage systems need to scale easily without time consuming or costly
upgrades.

Meanwhile, another key area is managing disk capacity that has a direct
bearing on reliability and availability as well. For instance, managing the same
in virtualized environments can be time consuming, complex and costly. Many
companies over-provision the storage for virtual machines as a result, and
depend on machine templates to allocate space, hoping that one size will fit
all. Also, setting up storage that must host multiple virtual servers can
require significant configuration to achieve optimal performance. The point here
is I/O bottlenecks can kill virtual server applications.

The third important thing is whether your management approach enables you to
get the best solution at the most economical TCO. So, reducing capital and
operating costs is an essential benefit of virtualization. But, many companies
under-utilize their storage capacity and spend little time researching that how
fast their storage really is in their virtualized environments. Companies need
storage systems that are easy to expand, yet deliver the best possible
performance value.

The design elements must also usher in the much needed flexibility.
Virtualization deployment is growing. More applications are moving from
traditional, physical servers to a virtual pool of resources. Storage is no
different. Companies need storage systems that scale as the virtualization
demand grows, without sacrificing the savings that virtualization provides.
Storage systems as an integral component of virtualization infrastructures
should easily scale, preserve virtual resources and adapt to changing
conditions. As a bottomline, the key thing one needs to factor in the design of
a virtualized tiered storage system starts with the applications. It is the
business needs and applications that drive the storage requirements which guide
tier configuration. Most applications can benefit from a mix of storage service
levels using high performance where it is important, and less expensive storage
where it is not.

Points to Ponder

Operationally it is not efficient to configure unique tiers for each
application. Individually configuring a unique scheme for each application leads
to extra work, cost and provisioning delays. Instead, the recommended practice
is to develop a catalog of pre-defined tiers with pre-defined characteristics,
and then allocate storage to applications as needed. Minimum upgrade steps
should be defined for all storage tiers to make sure that the defined
performance levels can be guaranteed at any time. The minimum upgrade step
should include multiple RAID groups to allow distribution of the new LUNs across
multiple drives and array groups.

The ultimate goal for any CIO is to put in place a tiered architecture that
can consolidate heterogeneous storage solutions into single manageable pools. It
eliminates fragmentation of the storage environment caused by disconnected
islands of storage and interoperability problems, and creates stranded capacity
and duplicates storage networking equipments, contributing to escalating
hardware costs. Eliminating barriers to sharing, storage recovery and improving
capacity utilization rates can deliver sustainable long term savings by allowing
future purchase of storage assets to be deferred.

Meanwhile, with heterogeneous configurations the norm in enterprise data
centers, managing storage requires the use of many software tools from different
vendors. These tools frequently do not communicate well with each other,
complicating the process of provisioning, optimizing, moving and protecting
data, and increases the training costs. Hence, to avoid heterogeneous storage
management its a good idea to migrate to a universal storage platform that can
weed out the inherent storage pain areas.

Shrikanth G

shrikanthg@cybermedia.co.in