Information Technology Reference
In-Depth Information
Transfer distance betwen two operations in
S
'
400
350
300
250
200
150
100
50
0
0
1
2
3
4
More
Transfer distances between nodes
Figure 8.17
Least transfer distance histogram of myExperiment workflows.
8.3.5 An Empirical Study
Given the matrix-based approach described in Section 8.3.4, we
calculate the least transfer distances and transfer paths between any
pair of service operations invoked in the samemyExperiment workflow
set obtained on August 23, 2010. As a reminder here, it contains
347 workflows, 241 services, and 283 operations. Figure 8.17 is the
histogram showing the least transfer distances between any pair of
nodes representing 283 operations in
S
0
,thatis,283
80,089
node pairs. 375 pairs of operations can reach each other without any
transfer, that is, they are connected within a single workflow. 147
operation pairs are reachable via 1 transfer; and 61 and 14 pairs are
reachable via 2 and 3 transfers, respectively. Only 2 pairs of operations
are reachable via 4 transfers; and there is no path with more than
4 transfers between any two operations.
In contrast to a public transportation system, the reachability
among the operations is obviously sparse: only 375
283
¼
þ
147
þ
61
þ
14
599 out of 80,089 operation pairs are reachable (about 0.75%);
if two nodes are not reachable within 4 transfers, they are not
reachable at all. The sparseness is due to two major reasons. First,
bioinformatics services/operations cannot be arbitrarily connected
because they have different usage scenarios by nature. Second, the
services in the myExperiment workflows largely function individually
rather than collaboratively, as discussed in Section 8.2.2. Although the
þ
2
¼
