Information Technology Reference
In-Depth Information
enough resources. Once the workload of grid resources exceeds a thresh-
old value, the grid system needs to scale up dynamically.
4.5.2.2
Fault Tolerance of Information Server
The information server fault-tolerant framework is shown in Figure 4.6. It
has three levels: distributed information server (DIS), grid information
index server (GIIS)- and resource provider. The Globus discovery service
records the identity and essential characteristics of “services” available to
domain members, and a GRIS registers its local resource information to a
GIIS in the same domain. Most of the time grid users access their domain
GIIS to get registered resources. They can also get local resource informa-
tion directly from a GRIS. Based on the distributed architecture and soft-
state registration, FTGP makes the system highly fault-tolerant.
In Figure 4.6, there are two GIISs in one domain. Server B is a backup of
server A, which records replicated categories. A grid notii cation frame-
work is added to GIIS. Server A is implemented as a notii cation source,
and server B is implemented as a notii cation sink. Server B subscribes to
server A as a listener. Once a new resource registers to server A or a reg-
istered resource leaves, server A will quickly notify such a change to
server B. Other information such as task results and the status of grid
applications will also be propagated to server B with the help of the grid
notii cation framework. In addition, server B detects server A to make
sure whether server A is valid or not. Once server A fails, server B will
Distributed information server
domain_1
domain_1N
domain_K
A
B
domain_11
domain_3
domain_2
register &
feedback
Registry node operator
Grid information index server
Resource provider
FIGURE 4.6
Information server fault-tolerant framework.
Search WWH ::
Custom Search