Storage Inefficiencies in Virtualized Environments
In highly virtualized environments, existing storage inefficiencies are
compounded. Why?
Overallocation: The tendency to allocate capacity based on
worst-case scenarios, called fat (or thick) provisioning, is done
to ensure that business units avoid capacity constraints. What
emerges, however, is a growing collection of underutilized storage
volumes, each far exceeding its actual requirements. Eventually, all
the capacity is provisioned, even though much of it isn’t used. The
result: an abundance of waste while at the same time a shortage of
capacity for new workloads.
Full-copy snapshots/cloning: The waste is multiplied whenever
fat-provisioned volumes are copied or archived on traditional arrays.
Redundant data is copied and archived, and unused space within
each volume is backed up as well, making offline storage a major
culprit in the misallocation of storage resources.
Performance degradation: An array of storage and networking
fabrics come into play when data moves to the cloud. And the sheer
physical distances covered can raise substantial concerns about
latency. How realistic is it to expect high-performance access to
storage resources that may be housed hundreds of miles away? And
assuming that high-performance technologies do exist, is the cost
justified for anything besides mission-critical applications?
Scale-out Storage: Intelligent Solutions
The explosion of digital content and the emergence of cloud
computing have driven the development of capacity-based scale-
out storage cloud architecture. These intelligent storage solutions
offer compelling benefits and a way for the next-generation data
center to address three major storage challenges: increasing volume,
inefficient management, and cost. While traditional approaches to
storage are built around extremely powerful, costly, and custom
hardware, intelligent storage solutions combine software and
converged storage servers to deliver high-speed access in a
much more modular and scalable solution. Multiple vendors are
now offering these innovative, cost-effective solutions based on
industry-standard x86 architecture.
There’s no one-size-fits-all solution for scale-out storage.
Depending on your organizational needs, you can design your
solution to support three basic usage models that are critical to
cloud computing:
Application data store: Data collected and stored as part
of business operations that needs to be available for further
manipulation (for example, data analysis)
Large object store: Photos, videos, thumbnails, and documents
in diverse formats stored and managed with applications such as
Microsoft SharePoint* and EMC* Documentum*
Backup and archiving: Scheduled backups that transfer data to
storage—typically a private cloud in most enterprises, but potentially
to software-as-a-service (SaaS) cloud storage, or even first to a
private cloud and then to the SaaS cloud
10