Special Features:
IDC REPORT ON GRIDS - NO
SIGNIFICANT NEW REVENUE!
IDC Research Report: Technical Cluster and Grid Taxonomy
IDC Analysts: Christopher G. Willard, Ph.D., Debra Goldfarb, Earl Joseph II,
Ph.D, and Nicholas J. Kaufman.
How do technical grids fit into the overall technical computing market? What
are the major drivers for this technology?
Technical grid-based computing strategies use system software, middleware, and
networking technologies to combine independent computers into logically
unified systems. The major distinguishing feature of grids is that they are
configured from components that are owned and/or managed by independent
individuals or organizations. Grids are largely composed of existing
resources, so they can be viewed as an operational concept or strategy rather
than a new class of computer systems products.
Advantages of the grid approach to distributed computing include increased
utilization of computing resources, access to specialized computer systems,
cost sharing, and improved management. The long-term market impact of grids
may center on the development of more complex computing infrastructures and
new mechanisms for accessing resources. Such evolutionary steps have
historically spurred overall growth in computer system utilization and
markets.
Grid Computing - An Idea Whose Time Has Come?
Grid technology has been under development since the late 1990s, with leading-edge users extending cluster and distributed computing concepts to allow
organizations to share and combine resources. In this case, wide area networks
plus software layers that help implement cross-organization resource
management policies allow users to create virtual computer systems that can
span organizations and geographies.
Over the last few years these concepts have begun to move into the broader
markets, with users looking to grid technology to optimize resources within
their organizations as well as support collaboration between organizations.
Standards are developing in the industry that will help support the broader
use and compatibility of grids.
Clustered Systems and Grid Definitions
For the purposes of our research, at the highest level IDC defines both
clusters and grids used in technical markets as a set of independent computers
combined into a unified system through systems software and networking
technologies.
Following are two general identifying features of technical clustered systems
and grids:
- Independent components. Clusters and grids are composed of complete
computer systems that could operate on their own outside of the cluster with minimal
additional modification
- Standard interconnects. Clusters and grids are generally connected via an
industry-standard technology, such as networking, I/O, or Web interfaces.
Note that both clusters and grids are approaches to distributed computing,
which represents the use of multiple computers combined to form a larger
computer system.
Separating Clusters from Grids
We divide cluster and grid technology based on whether or not there is an
organization component in configuring the system. Thus, the key differences
are as follows:
- Clusters use dedicated components. All components in a cluster are
exclusively "owned" and managed as part of the cluster. All resources are know
and fixed, and there is a dedicated interconnect used between the nodes.
- Grids share resources among independent owners. Grids are configured from
computer systems that are generally managed and used both as part of the grid
and as independent systems. Thus, the individual components of the grid are
not fixed in the grid, and the overall configuration of a grid changes over
time. In some case only the control software will have an accurate accounting
of resources at a given moment.
Major characteristics of grids include the following:
- Organizational component -- The major distinguishing feature of grids is
that they are configured from components that are owned and/or managed by
independent individuals or organizations. This ownership ranges from
individual engineers who can disconnect or turn off their workstation at night
to avoid it being used as part of a group-level batch processing resource
during off hours to a university or research center that will grant specific
access rights to other organizations as part of a resource and/or data-sharing
agreement.
Thus, grids have an organization component that defines such system
characteristics as:
- The components that are part of the grid
- How different components can be accessed
- What resources each member organizations can use and when
- Virtual Systems -- One result of the organizational component of grids is
that grids are dynamic complexes of systems in which component resources can
change at any time. Thus these complexes can be viewed as forming a "virtual"
computer. From the user's perspective resources are virtualized so that the
specific resources used to run a program or to store data can become invisible
to the user. In contrast, cluster users work with a specific set of well-
defined resources.
In theory, this virtualization of resources leads to a utility model of
computing where the sources of computing cycles become irrelevant to end
users. In this case, resources are brokered by a grid utility, which also
handles accounting and charge back.
- Dynamic configuration -- The physical structure of the grid is defined in
part by the organizational structure, and thus resources can be added or
removed from the grid at any time. Such changes can be regularly scheduled
events, the results of long-term planning or contracts, or virtually random
occurrences.
- Geographic distribution -- Multiple owners imply multiple locations for
equipment. Geographic distribution of components can range from interoffice to
national to global.
Grid Objectives: A Second Dimension in Defining Grids
Grids can also be categorized by primary use or design objectives. We consider
three broad objective classes, as follows:
- Compute grids. These grids share access to high-performance computer
systems.
- Data grids. These grids share access to databases and file systems.
- Utility grids. These grids share access to application software,
computing, and other resources.
It is important to note that individual grids at any point along the
complexity spectrum can in theory fall into one or more of the classes. That
said, utility grids present commercial opportunities and thus may be more
heavily represented at the high end of the complexity spectrum (i.e.,
provisioning grids).
Market Impacts and Sizing Issues of Grids
IDC views grids as operational concepts rather than new computer hardware
products. That is grid technology provides new approaches to organizing and
accessing computer resources rather than new types of computer systems. In
addition, grid strategies are generally used for managing existing resources,
and thus grids in and of them selves are not associated with new systems
sales.
In a worst-case scenario grids can act as market dampers by increasing the
utilization efficiencies of existing resources and thus extending buying
cycles. Efficiencies are gained through greater access to idle computers and
reduced duplication of resources. It should be noted that this scenario
assumes relatively low utilization rates of computer systems. Low utilizations
is often the case for desktop systems, but it is generally not the case for
high-performance servers.
Measuring Grid Market
In estimating the size of the "grid computer market," IDC must avoid double
counting of computer sales. IDC currently sees the grid market as a new use
for computers already accounted for in market measurements. In one view, the
grid market is a horizontal cut of the current computer markets, looking at
which computers are also used as part of a grid.
Given that grids will not directly generate significant new revenue in the
technical market, then the questions arises about how to measure the current
use and future growth of grid technology. IDC believes there are three basic
approach to measuring grids:
- Track grid-enabling software. This would determine the number of grid
software licenses sold or provided through open source. This approach would
count software license acquisition but not necessarily whether the software
was running on a grid system. In addition, it would not provide information on
the number of grids in use, nor how they were configured, and so on.
- Survey enduser organizations on grid implementations. This approach would
provide information on grid configurations, applications, expected growth
patterns, and so forth. Data from such a survey could be used to estimate
overall market sizes.
- Seek out grid-based services businesses. Identify organizations that sell
technical solutions based on grid technologies. Survey these organizations to
track market size and growth characteristics for businesses based on grid
technology or concepts.
IDC is currently evaluating these approaches for measuring grid-based markets
in terms of both feasibility and the value of results to our customers.
Conclusion
IDC believes that grid computing can be viewed as part of the migration of
computing technology into the overall scientific research and engineering
development infrastructures. Grids allow the implementation of resources
management and parallel programming strategies across organization and
geographic boundaries. The advantages of grid approaches include increased
utilization of computing resources, access to specialized computer systems,
cost sharing, and improved management. Long-term markets impacts of grids may
center around the development of more complex computing infrastructures and
new mechanisms for accessing resources.
http://www.idc.com
http://www.idc.com/hpc
|
