China Grid and Related Dependability Research - Grid Computing: Infrastructure, Service, and Applications - page 87

following sections: ChinaGrid SuperVison, a national-wide grid monitoring
system; ALTER, an adaptive grid fault detector [34]; FTGP, a fault-tolerant
grid architecture [35]; and DRIC, an adaptive application fault-tolerance
framework.

4.5.1 Grid Monitor

Monitoring resources is an important aspect of the overall efi cient usage
and control of any distributed system, and for improving the dependabil-
ity of the whole system. The primary idea of grid monitoring is to integrate
heterogeneous monitoring deployments or facilities of grid participants. In
the ChinaGrid, a nationwide grid monitoring system has been developed
and deployed, called ChinaGrid SuperVison (CGSV).
CGSV, sponsored by HP, is a program launched to handle monitoring-
related issues for the ChinaGrid project. It was exhibited at the China
National Computer Conference (CNCC) 2005. In accordance with the lat-
est efforts in SOA and Grid communities, for example, WSRF and WSDM
[36], CGSV seeks a set of resource measurement and control services for
the operation of the ChinaGrid ecology. CGSV tries to provide system sta-
tus information for CGSP components and application users. Moreover,
CGSV tries to provide system trace data for analysis, to improve the system
dependability and resource usability. The difference between CGSV and
the CGSP Information Center is that CGSV puts more effort into dynamic
information while the Information Center focuses more on static
information.

4.5.1.1 Architecture

CGSV i s desig ned based on t he g r id mon itor i ng a rc h itect u re [37] proposed
by the Global Grid Forum. As shown in Figure 4.4 , CGSV is built up by
several components, which can be divided into three layers: the collection
layer, service layer, and presentation layer. At the collection layer, sensors
are deployed for data collection. Sensor-I is responsible for collecting
hardware resource information. Other sources of information, including
network condition and dynamic service information or data generated by
other monitoring tools, are wrapped to unify the data format by an adap-
tor. Above the data interface, based on a transfer and control protocol, lies
the service layer. In each domain of the ChinaGrid, this layer is presented
as a logical domain monitor center, where the domain registry service and
monitor services are deployed. A message gateway service is provided to
alleviate the communication cost of monitor services since over-notii cation
and over-subscription will disastrously decease service performance. The
top layer is the presentation layer, where data analysis, visualization, and
management can be performed. A detailed description of sensors, proto-
col and stream-integration design will be given in the next section.