Hardware Reference
In-Depth Information
To this end, grid researchers are developing services, tools, and protocols to enable
these virtual organizations to function.
The grid is inherently multilateral with many participants who are peers. It can
be contrasted with existing computing frameworks. In the client-server model, a
transaction involves two parties: the server, who offers some service, and the client,
who wants to use the service. A typical example of the client-server model is the
Web, in which users go to Web servers to find information. The grid also differs
from peer-to-peer applications, in which pairs of individuals exchange files. Email
is a common example of a peer-to-peer application. Because the grid is different
from these models, it requires new protocols and technology.
The grid needs access to a wide variety of resources. Each resource has a spe-
cific system and organization that owns it and that decides how much of the re-
source to make available to the grid, during which hours, and to whom.
In an
abstract sense, what the grid is about is resource access and management.
One way to model the grid is the layered hierarchy of Fig. 8-52. The
fabric
layer
at the bottom is the set of components from which the grid is built. It in-
cludes CPUs, disks, networks, and sensors on the hardware side, and programs and
data on the software side. These are the resources that the grid makes available in
a controlled way.
Layer
Function
Applications that share managed
resources in controlled ways
Application
Discovery, brokering, monitoring
and control of resource groups
Collective
Secure, managed access
to individual resources
Resource
Physical resources: computers, storage,
networks, sensors, programs and data
Fabric
Figure 8-52.
The grid layers.
One level higher is the
resource layer
, which manages the individual re-
sources. In many cases, a resource participating in the grid has a local process that
manages it and allows controlled access to it by remote users. This layer provides
a uniform interface to higher layers for inquiring about the characteristics and stat-
us of individual resources, monitoring them, and using them in a secure way.
Next is the
collective layer
, which handles groups of resources. One of its
functions is resource discovery, by which a user can locate available CPU cycles,
disk space, or specific data. The collective layer may maintain directories or other
databases to provide this information. It may also offer a brokering service by
which the providers and users of services are matched up, possibly allocating
scarce resources among competing users. The collective layer is also responsible
