Information Technology Reference
In-Depth Information
take over the work of server A. The registered information includes host
information (such as static and dynamic information about CPU, memory,
disk, and system software), task results, and applications that provide
specii c services for i nal users.
By introducing DIS to the Globus information server framework, FTGP
provides a convenient means for grid users to search all the grid resources.
DIS is a distributed system, which is internally constructed as a peer-to-
peer [44,45] system. DIS consists of several registry node operators.
Information registered to each registry node operator is regularly copied
to the other node operator; that is, information is replicated among internal
nodes. The architecture of the DIS works just like distributed UDDI. The
GIIS interacts with the DIS with three API types: registered API, updated
API, and terminated API. An active GIIS summarizes its registered
resources and application services and reports them to the DIS. Grid users
can get more resources and service information of other communities
from the DIS. The DIS is also implemented as a notii cation sink, and
accepts the resource changes from the active GIISs.
4.5.2.3
Fault Tolerance of the System-Level Component
There are complicated relationships among system-level components.
Any component failure may lead to system crash. The local monitor is also
a system-level component. When other components start, the processes
register to the local monitor, and submit a handling script simultaneously.
The handling script dei nes their dependent relationship and their startup
order. Once the local monitor detects that any process fails, it will recover
that process from failure according to the handling script. FTGP provides
a self-maintenance mechanism for the local monitor. There are two local
monitors in the system: the active and backup. There is a shared slot table
used to record registered processes. When the active one fails, the backup
one will take over the work according to the slot table, and restart another
backup local monitor.
4.5.2.4
Task Fault Tolerance
A grid resource broker (GRB), such as Nimrod/G [46], provides a graphical
interface to monitor and utilize grid resources [47]. GRB records information
about user tasks and scheduled information. Information service [48] is a
key part of grid software infrastructure, providing fundamental mecha-
nisms of discovery and monitoring. Globus Information Services consists
of a GIIS and a grid resource information server (GRIS). User tasks are
submitted to GRB, which searches resource information from GIIS and
schedules user tasks to the suitable grid node. Local components in grid
nodes register to GIIS to publish their resource information and locations
by lightweight directory access protocol (LDAP). GRIS is responsible for
Search WWH ::
Custom Search