Cloud on demand

While the notion is true that cloud infrastructures is poised to
revolutionize the way we access technology, it is also true that it is a vast
area and enterprises often grapple with understanding its full potential. The
ever increasing complexity of data centers coupled with the massive growth of
data is leading to an increase in resources required to store, process, optimize
and serve information back to end users when required. Organizations have
rapidly adopted virtualization and are now looking at cloud delivery service
models to reduce costs, increase flexibility and improve their time to market.

To keep pace with today's competitive marketplace, enterprises require
greater flexibility and agility than their traditional storage architectures can
provide. This, coupled with explosive unstructured content growth, is driving
organizations to look toward instant IT delivery models.

On-demand IT Delivery

In the move towards on-demand IT delivery, it hasn't taken long for
customers to recognize the benefits of private, hybrid or public cloud models,
the most tangible being significant cost savings. From a capital expenditure
perspective, organizations tend to over-purchase to deal with the ebb and flow
of storage and resource requirements to support the business. This often leaves
them with an abundance of underutilized hardware assets. The ability of cloud
infrastructures to grow and contract storage resources, in tune with the
business needs, minimizes this upfront capital expense, moving them from 'fixed
costs' to 'variable costs'.

If we take a look at the operational costs, 'hidden costs', the data growth
and complexity to manage traditional IT environments has emerged. Customers can
reduce much of this operational expenditure by deploying cloud models and
leveraging cloud managed services, paying only for what they consume and
eliminating the day-to-day management tasks altogether.

While on-demand access to computing resources is what the industry is
striving for, it also poses concern and risk for IT organizations. This can be
very disruptive to business process and control. If business users begin to
outsource to cloud providers in order to get faster support, sensitive
information could be put at risk. To mitigate this risk, IT organizations should
be thinking about developing an internal cloud enabled architecture to provide
greater business agility, in addition to a process for ad hoc projects that
require to be outsourced to a public or hybrid cloud provider. This way they can
move into the cloud in a controlled and orderly manner mitigating associated
risks.

There are many deployment choices to consider for moving into the cloud.

Private cloud: For simplicity, let's define a private cloud as cloud enabled
infrastructure within the physical walls of a data center. A private cloud can
provide many of the benefits of cloud without the security risks associated with
public deployments. Because it is accessed over an internal network or intranet,
it's as secure as the rest of the data. Since one controls it and the
environment around it (ie networks, servers, etc), one can achieve enterprise
level SLAs. But one do sacrifice some of the operational cost savings such as
physical floor space, power, and cooling. Unless one is leveraging a managed
service one is also subject to management overhead.

Hybrid cloud: Now let's take a look at the hybrid or trusted cloud, which we
will define as infrastructure that resides at a trusted service provider. In
this case, access is limited to appropriate resources in one's organization and
delivered over a virtual private network or a secure Internet connection. Since
the infrastructure is out of the organization's direct control, service levels
could be impacted by external factors. Customers also need to think about the
physical security of the environment, which is why it is important to understand
the service provider's processes and requirements around physical access.

Public cloud: Lastly, the public cloud can be described similarly to the
hybrid, except that there is usually more general access over the Internet
providing limited security. Many public cloud offerings are quite inexpensive or
sometimes even free and SLAs are generally not guaranteed or measured
differently than how an enterprise measures their SLAs. Additionally, value
added services and features such as encryption, compression, back-up, tiering
and replication are not available from public providers as they are from private
or hybrid cloud providers.

What Makes a Cloud?

Regardless of the type of cloud, there are some key features every cloud
platform should have. Firstly, it is a secure, direct connection to get data
into the cloud, such as a representational state transfer (REST) interface or an
on-ramp to connect applications to the cloud without requiring application
recoding. REST is an approach for getting information content from a website by
reading a designated web page that contains an Extensible Markup Language (XML)
file that describes and includes the desired content. There also needs to be
multitenancy capabilities to logically segregate the data, so that SLAs can be
assigned to specific data types or applications. The cloud should also have
namespaces with access rights and security layers to prevent unauthorized
access. Depending on the provider, some clouds offer value added features like
compression and single instancing to improve cost savings, encryption to provide
greater security and billing and chargeback for organizations or service
providers that wish to bill each business unit or organization based on
consumption.

IT services are generally held to service level standards for availability,
reliability and integrity. Augmenting or replacing 'legacy' IT services with
cloud services require the same quality guarantees. Although, managed cloud
services allow customers to focus less on the storage management it is critical
that customers include an expected quality of service in their contracts with
cloud providers. Not all cloud providers measure SLAs the same way, so it's
important to ask how they quantify their SLAs.

Given some of the trade-offs between the various cloud deployment models, how
does one identify the most appropriate candidates for deployment? One should
start by identifying the data in one's environment that generally has lower
business value and lower SLA requirements. For example, think of data types like
home directory shares, static data or backup content that can be moved from
onsite 'primary' to cloud 'secondary' storage.

One can get immediate cost savings by moving this peripheral data-data that
doesn't require active management or constant read/write access to the cloud. It
is unnecessary to pay such high administrative and management overhead for this
non-business critical data. frees up resources to focus on the core business
applications, improving operational efficiency and utilization of one's existing
assets. Next, it allows one's organization to gain experience and develop best
practices for cloud deployments. Lastly, it allows one to move toward the core,
tier-1 applications at one's own pace.

Sunil Chavan

The author is Director, Software Group & Cloud Solutions, Asia Pacific,
Hitachi Data System

(Source: DQ)