Information Technology Reference
In-Depth Information
VO
Back-up
index
Main
index
Index
node
Index
node
Site B
Index
node
Index
node
Index
node
Index
node
Index
node
Index
node
Index
node
Index
node
Index
node
Index
node
Site A
FIGURE 13.2
Hierarchical MDS4 structure in a VO.
of these failures occur at a main index service, it can cause the index
service network to break into isolated subnets requiring manual mainte-
nance. MDS4 also requires that its structure as shown in Figure 13.2
be predei ned on container startup, imposing a i xed and nondynamic
architecture that will unquestionably fail if the information service net-
work experiences any form of churn . Messages l owing up or down along
the hierarchical structure will also experience delays as trafi c involved in
keeping an up-to-date index increases.
An index service is generally coni gured to update its data at predei ned
intervals, which can be large (e.g., 5 minutes) in some cases. This can cause
havoc with any form of time-sensitive services. On the other hand,
reducing these intervals could create an expected increase in trafi c and
processing usage. To increase the performance of MDS4, the associated
hierarchical structure must be replaced with a more scalable, dynamic,
and robust structure.
13.3
The main goal of PIndex is to enhance information services for grid
environments through P2P techniques. However, it has been revealed
by Foster and Iamnitchi [10] that a pure P2P approach cannot operate
PIndex for Scalable Grid Information Services
 
 
Search WWH ::




Custom Search