 |
|
DAILY NEWS AND INFORMATION
FOR THE GLOBAL GRID COMMUNITY /
|
Special Features:
ORACLE'S SOUDER SPEAKS GRID
REALITIES, POSSIBILITIES
The following is a summary of views expressed by Benny Souder, vice
president
of distributed database development at Oracle Corp, in Line56, an e-business
ezine.
Grid computing is the notion of computing as a utility. We get what we need
without tradeoffs. With Grid computing, the computer's location is irrelevant
-- computation is available when needed. When we ask for data, we get as much
as we need. That is the vision: Computing and data that are ubiquitous and
always available.
On the client side, it's simple. Computing power is readily available
through
a standard interface. However, behind the scenes on the server side, it's
pretty sophisticated, with several essential components needed within the
infrastructure and architecture to make this happen.
Grid computing is based on two components. "Virtualization" means severing
the
hard coded association of resources to systems and provisioning means, making
resources available whenever and wherever needed. Instead of attaching disks
to each computer, storage can be virtualized into a central pool and allocated
to computers when they need it.
Grid computing also is about consolidation. Grid is distributed
consolidation
-- a smaller number of larger resource pools and processing centers -- and
dynamically allocating resources as needed.
Dynamically allocating allows computing to align with the needs of the
business and stay aligned as needs and priorities change. We don't expect to
have to adapt to a service like electric power, it adapts to us. Grid
computing adapts to and dynamically aligns with your business.
The advantages don't stop with this dynamic alignment, however. Most
applications are independently constructed, custom configured and sized for
peak load -- which may occur only once a year -- leaving resources
underutilized most of the time.
This is accepted as a tradeoff and forces people to choose between having
enough scalability for peak, and not buying too much so that they've invested
a lot of money in idle capacity. This is not the case with Grid computing.
Grid computing allows users to allocate more servers to wherever it is
necessary at that time. When not needed for one task, the power can be
utilized for others.
Grid computing represents the next wave of standardization. Networks have
standardized. Intel and AMD processors have become the standard. Interconnects
are rapidly moving to standards such as Infiniband. Standard operating systems
are starting to emerge.
Standardizing on hardware makes it easier to re-deploy when needs change or
to
re-purpose when the business changes. Standardizing on software requires
fewer, more widely applicable skills.
Against the backdrop of standardization, software infrastructures capable
of
dynamic provisioning begin to emerge. Almost every infrastructure provider
envisions the virtualization and provisioning that is Grid computing. The
visions of most providers are all aligned around Grid computing and the
smaller number of standard components that they must support will accelerate
these infrastructures.
Standard components will build the best Grids. Small towers or server
blades
deliver the fastest performance and the lowest cost. They can be two or three
times faster than larger computers and half, to a quarter, of the cost. Since
they are smaller than large computers, they constitute a more fine-grained
unit of allocation, allowing Grid technologies to allocate exactly the amount
of resource needed with no waste. Therefore, the most powerful Grids will be
built with the lowest-cost hardware. Businesses will save money, get faster
computers and better allocation for better utilization and responsiveness.
"Scale out" is another basic component of Grid computing based on the idea
that we can start small and then simply add more components as we scale. This
means that Grid computing can happen incrementally. Organizations can start
small, gain experience, and as they are successful, continue to move
applications.
A large number of organizations are now participating in Grid standards,
building Grid pilots, or getting ready to deploy Grid solutions. Since Grid is
incremental and standard, many organizations actually are doing things right
in line with Grid computing without even realizing it.
Grid computing has obvious advantages for large organizations. It is easy
to
see how they can pool these resources and then allocate them as needed, and
re-allocate them when their needs change.
However, Grid computing also offers benefits to smaller organizations.
Medium-sized organizations likely need enough computing to build useful pools
of resources. In some of these organizations, the peaks can actually be a
larger percentage of the total computational work, making dynamic allocation
even more valuable. For this reason the incremental investment to get to Grid
computing will be very compelling for these organizations.
Outsourcing to a service provider that can make allocation decisions for
your
business is a good idea. The small business gains resources needed at peak
without having to invest in them on a permanently.
Capability and standardization are about to wash over us -- it is
inevitable
and unstoppable. The best advice is to gain early advantage and reap the
benefits.
|